docs(v3): Hadamard-residual-ladder honourable mention next to the three flavours of 256 + Hambly–Lyons [S] anchor#700
Conversation
…esidual ladder next to the three flavours of 256; Hambly-Lyons [S] anchor Operator-directed. Adds to le-contract.md §3 (append-only subsection, no change to the L1-L8 catalogue): the index+residual fourth operational mode — bgz-tensor AdaptiveRow keeps the palette/CLAM centroid as the deterministic PLACE and stores only a Hadamard-rotated residual (i8 outlier rows / i4+i2 cascade regular rows). It is the continuous-field exit flat L4 cannot provide (geo-arc instance: 256 levels over ~1500 m relief = ~6 m elevation terraces); categorical surfaces (Signed360 normals, narrow colour) stay flat L4. NOT a ninth 12-byte layout — the ladder is out-of-row (Signed360 status); in-row refinement remains the turbovec 6x4-bit nibble lane. Formal anchor graded [S]: Hambly-Lyons 2010 (Annals of Math 171) signature uniqueness — a bounded-variation path is determined by its graded iterated-integral cascade up to tree-like equivalence; the theorem-shaped version of the ladder's promise. Stays [S] until a ladder->signature probe exists; [FORMAL-SCAFFOLD] consult required for promotion. Board hygiene in the same commit: EPIPHANIES E-PALETTE-RESIDUAL-LADDER-1 prepended; LATEST_STATE v4 entry prepended; CLAUDE.md's stale "bgz-tensor 0 deps" row corrected (it path-deps ndarray + holograph — verified against crates/bgz-tensor/Cargo.toml). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: serverGenReqId_48ea823f-c9a2-4f1c-9e59-04476e5e370a) |
|
Warning Review limit reached
Next review available in: 20 minutes 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)
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 |
…ay hpc pillar B7; only the ladder->signature mapping stays [S] Operator pointer: the signature theorem is not an external citation. The theorem side ships in BOTH repos: - jc Pillar 11 (crates/jc/src/hambly_lyons.rs, feature hambly-lyons -> sigker): forward probe = tree-equivalent excursion collapses to the identity signature via Chen's identity; converse probe = triangle's non-zero level-2 signed area; deliberately on signature_truncated, not the known-buggy signature_kernel_pde. - ndarray src/hpc/pillar/signature.rs (B7): signature transform + Hambly-Lyons sig-kernel, Gram-matrix PSD check, certification probe over 1,000 Levy paths. The honourable-mention text previously said the anchor was "unproven in this workspace" — corrected: only the ladder-levels -> signature-levels MAPPING for the residual cascade stays [S], and its probe can build on the existing Pillar-11 harnesses. le-contract subsection, EPIPHANIES entry, and LATEST_STATE entry updated consistently (all in-flight text of this PR, not settled history). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki
![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: a0868a91fa
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".
| `AdaptiveRow { centroid_idx: u16, scale_bf16, scale2_bf16, … }` keeps the | ||
| palette/CLAM centroid as the coarse deterministic PLACE and stores only a | ||
| **Hadamard-rotated residual**, quantized i8 (outlier rows, ~top 30%) or as | ||
| an i4+i2 cascade (regular rows, ~bottom 70%). Index PLACES, residual |
There was a problem hiding this comment.
In the current crates/bgz-tensor/src/adaptive_codec.rs, classify_rows_by_lfd sends the top 10% LFD rows to RowPrecision::Passthrough, and those rows store the original vector in passthrough with no centroid/residual payload; only the next 20% are i8 and the bottom 70% are i4+i2. Describing all outlier rows as the ~top 30% i8 residual tier makes the new contract text false for the hardest rows and can mislead consumers that rely on this section to decide whether every row has an index+residual representation.
Useful? React with 👍 / 👎.
…% / i8 20% / i4+i2 70% (codex P2) classify_rows_by_lfd (adaptive_codec.rs, verified): the hardest ~top-10% LFD rows go to RowPrecision::Passthrough — the exact original vector is stored with NO centroid/residual representation — then ~20% i8 residual, ~70% i4+i2 cascade. The earlier "~top 30% -> i8" wording was false for the hardest rows; consumers must not assume every row has an index+residual representation. le-contract subsection + EPIPHANIES entry aligned (both in-flight text of this PR). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki
…/helix; TD-BGZ-TENSOR-PRE-LANE-REVIEW queued Operator nudge: the adaptive codec's mechanisms (Hadamard rotation, bespoke i4+i2 cascade, Passthrough escape) were engineered before today's lane inventory existed. The honourable mention records what SHIPS, not a certification of current best practice. - TECH_DEBT: new TD-BGZ-TENSOR-PRE-LANE-REVIEW entry — engineering follow-up review scoped: (a) i4+i2 cascade vs turbovec Lloyd-Max + NativeLut; (b) Hadamard vs the PolarQuant rotation-vs-error-correction findings; (c) residual/location coding vs helix ResidueEdge/Signed360 + CurveRuler; (d) Passthrough routing to an existing full-precision lane. Outcomes: consume-the-lane / demarcate / retire. - le-contract honourable mention + EPIPHANIES entry carry the caveat and the TD pointer (in-flight text of this PR). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki
2 participants
Summary
Operator-directed follow-up to #699 (branch restarted from main). Adds the index + residual fourth operational mode to the V3 substrate docs, next to the three flavours of 256:
le-contract.md §3— new append-only "Honourable mention" subsection (the L1–L8 catalogue itself is untouched): bgz-tensor'sAdaptiveRow(adaptive_codec.rs, verified against source) keeps the palette/CLAM centroid as the coarse deterministic PLACE and stores a Hadamard-rotated residual in a three-tier LFD split (corrected per codex P2,classify_rows_by_lfd): the hardest ~top-10% LFD rows escape toPassthrough(exact original stored — no index+residual representation, the codec's own refuse-to-force-the-mold tier), next ~20% → i8 residual, bottom ~70% → i4+i2 cascade. It is the continuous-field exit flat L4 cannot provide (geo-arc instance: 256 levels over ~1500 m relief ≈ 6 m elevation terraces; categorical surfaces likeSigned360normals stay flat L4). Explicitly NOT a ninth 12-byte layout: the ladder is out-of-row (same status asSigned360in the 96-bit carving); in-row refinement remains the turbovec 6×4-bit nibble lane.crates/jc/src/hambly_lyons.rs, featurehambly-lyons→sigker; forward probe = Chen-identity tree-collapse, converse = triangle level-2 signed area, onsignature_truncatednot the known-buggysignature_kernel_pde) and ndarraysrc/hpc/pillar/signature.rs(B7: sig transform + sig-kernel, Gram PSD check, 1,000-Lévy-path certification probe). Only the ladder→signature mapping stays [S] ([FORMAL-SCAFFOLD] consult required for promotion); its probe builds on the existing Pillar-11 harnesses.E-PALETTE-RESIDUAL-LADDER-1prepended; LATEST_STATE v4 entry prepended.ndarray+holograph(verified againstcrates/bgz-tensor/Cargo.toml).Companion ndarray PR #244 mirrors both corrections into the merged §8 flavours block.
Review
67fb9c7, verified againstadaptive_codec.rs:97-135.Verification
adaptive_codec.rs(AdaptiveRow fields,hadamard_rotate,classify_rows_by_lfd10/20/70 split),attention.rs(65,536-score / 128 KB table),Cargo.toml(deps),jc/src/hambly_lyons.rs+hpc/pillar/signature.rs(Pillar-11 implementations).🤖 Generated with Claude Code
https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki