From 5d2767f172daaf31780783e157282639ef60ecce Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 16 Jul 2026 08:25:50 +0000 Subject: [PATCH 1/3] =?UTF-8?q?docs:=20Hambly-Lyons=20anchor=20is=20in-wor?= =?UTF-8?q?kspace=20=E2=80=94=20Pillar-11=20B7=20here=20+=20jc=20Pillar=20?= =?UTF-8?q?11;=20only=20the=20ladder->signature=20mapping=20stays=20[S]?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Follow-up to merged #243 (branch restarted from master). The §8 fourth-mode sentence said the Hambly-Lyons anchor was analogy-grade "until a ladder->signature probe exists" without noting the theorem side already ships in both repos: - THIS repo: src/hpc/pillar/signature.rs (Pillar-11 B7 — signature transform + Hambly-Lyons sig-kernel, Gram-matrix PSD check, certification probe over 1,000 Levy paths). - lance-graph: crates/jc/src/hambly_lyons.rs (Pillar 11, feature hambly-lyons -> sigker; forward/converse probes on signature_truncated). Amended so only the ladder-levels -> signature-levels MAPPING for the residual cascade stays [S], with its probe building on the existing Pillar-11 harnesses. Blackboard note appended. Mirrors lance-graph ec58954 (PR #700). Knowledge-doc suite green (117/117). Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki --- .claude/blackboard.md | 6 ++++++ .../knowledge/pr-x12-h268-morton-wgpu-synergies.md | 12 +++++++++--- 2 files changed, 15 insertions(+), 3 deletions(-) diff --git a/.claude/blackboard.md b/.claude/blackboard.md index 06386f28..a59e666f 100644 --- a/.claude/blackboard.md +++ b/.claude/blackboard.md @@ -769,3 +769,9 @@ default is x86-64-v3 (avx2) so ndarray_simd activates on avx512 builds only. probe], cache-native 192B working set) — all nine consequences stay **probe-gated** (D-MTS-1..3, PHASE-1/PERT-RHO/PYR-1, WHP-1..4, L4 doc-lock); no kill condition weakened. +- **§8 fourth-mode follow-up (post-#243):** the Hambly–Lyons anchor is + in-workspace, not external — THIS repo `src/hpc/pillar/signature.rs` + (Pillar-11 B7: sig transform + sig-kernel, Gram PSD, 1000-Lévy-path + certification) + lance-graph jc Pillar 11 (`hambly_lyons.rs`, feature + `hambly-lyons` → sigker). Only the ladder→signature MAPPING stays + [S]; probe builds on the Pillar-11 harnesses. §8 sentence amended. diff --git a/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md b/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md index 25e1395a..e147b93d 100644 --- a/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md +++ b/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md @@ -249,10 +249,16 @@ rows) — index PLACES, residual CORRECTS. It is the continuous-field exit where a bare 256-level index terraces (e.g. elevation); categorical surfaces (`Signed360` normals, narrow colour) stay flat. Out-of-row like `Signed360`; the in-row refinement budget remains the turbovec nibble -lane. Formal anchor [S]: Hambly–Lyons 2010 signature uniqueness (a +lane. Formal anchor: Hambly–Lyons 2010 signature uniqueness (a bounded-variation path is determined by its graded iterated-integral -cascade up to tree-like equivalence) — analogy-grade until a -ladder→signature probe exists. Canonical text: lance-graph +cascade up to tree-like equivalence) — **already in-workspace on both +sides** (corrected 2026-07-16): THIS repo's `src/hpc/pillar/signature.rs` +(Pillar-11 B7 — signature transform + sig-kernel, Gram PSD check, +1 000-Lévy-path certification probe) and lance-graph jc Pillar 11 +(`crates/jc/src/hambly_lyons.rs`, feature `hambly-lyons` → `sigker`, +forward/converse probes on `signature_truncated`). Only the +ladder→signature MAPPING for the residual cascade stays [S]; its probe +builds on the existing Pillar-11 harnesses. Canonical text: lance-graph `le-contract.md §3` honourable-mention subsection + `E-PALETTE-RESIDUAL-LADDER-1`. From f0ca32b87972ee81c3d952c2c2c564b7009f0269 Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 16 Jul 2026 08:27:10 +0000 Subject: [PATCH 2/3] =?UTF-8?q?docs:=20correct=20residual-ladder=20tier=20?= =?UTF-8?q?split=20in=20=C2=A78=20=E2=80=94=20Passthrough=20top-10%=20/=20?= =?UTF-8?q?i8=2020%=20/=20i4+i2=2070%?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Mirrors the codex P2 fix on lance-graph #700: classify_rows_by_lfd sends the hardest ~top-10% LFD rows to Passthrough (exact original stored, no index+residual representation), next ~20% to i8, bottom ~70% to i4+i2. The earlier "i8 for outlier rows" wording was false for the hardest rows. Knowledge-doc suite green (117/117). Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki --- .claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md b/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md index e147b93d..808032db 100644 --- a/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md +++ b/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md @@ -243,9 +243,13 @@ the arithmetic, a 256² u16 table is 128 KB, and the attribution.) **Honourable fourth mode (operational, not a footprint): index + residual.** bgz-tensor's `AdaptiveRow` (`adaptive_codec.rs`) keeps the -palette centroid index as the coarse deterministic PLACE and stores only a -Hadamard-rotated residual (i8 for outlier rows; i4+i2 cascade for regular -rows) — index PLACES, residual CORRECTS. It is the continuous-field exit +palette centroid index as the coarse deterministic PLACE and stores a +Hadamard-rotated residual in a three-tier LFD split +(`classify_rows_by_lfd`): the hardest ~top-10% LFD rows escape to +`Passthrough` (exact original stored — no index+residual at all), the +next ~20% get an i8 residual, the bottom ~70% the i4+i2 cascade — index +PLACES, residual CORRECTS, and the mold is never forced on rows it +doesn't fit. It is the continuous-field exit where a bare 256-level index terraces (e.g. elevation); categorical surfaces (`Signed360` normals, narrow colour) stay flat. Out-of-row like `Signed360`; the in-row refinement budget remains the turbovec nibble From d1daecbd9a482202c140e23d64adc75b7617e9fa Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 16 Jul 2026 08:32:35 +0000 Subject: [PATCH 3/3] =?UTF-8?q?docs:=20provenance=20caveat=20on=20the=20?= =?UTF-8?q?=C2=A78=20fourth=20mode=20=E2=80=94=20bgz-tensor=20predates=20t?= =?UTF-8?q?urbovec/PolarQuant/helix?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Mirrors lance-graph 88dcf6b: the fourth-mode paragraph now notes the adaptive codec was engineered before today's lane inventory existed and points at the queued engineering follow-up review (TD-BGZ-TENSOR-PRE-LANE-REVIEW: i4+i2 cascade vs turbovec Lloyd-Max/NativeLut, Hadamard vs PolarQuant rotation findings, residual coding vs helix — consume-the-lane / demarcate / retire). Knowledge-doc suite green (117/117). Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki --- .claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md b/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md index 808032db..fee7c304 100644 --- a/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md +++ b/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md @@ -249,7 +249,12 @@ Hadamard-rotated residual in a three-tier LFD split `Passthrough` (exact original stored — no index+residual at all), the next ~20% get an i8 residual, the bottom ~70% the i4+i2 cascade — index PLACES, residual CORRECTS, and the mold is never forced on rows it -doesn't fit. It is the continuous-field exit +doesn't fit. Provenance caveat: bgz-tensor predates turbovec/PolarQuant +and helix — this records what ships, not current best practice; an +engineering follow-up review is queued (lance-graph +`TD-BGZ-TENSOR-PRE-LANE-REVIEW`: reconcile the i4+i2 cascade vs turbovec +Lloyd-Max/NativeLut, Hadamard vs the PolarQuant rotation findings, +residual coding vs helix — consume-the-lane / demarcate / retire). It is the continuous-field exit where a bare 256-level index terraces (e.g. elevation); categorical surfaces (`Signed360` normals, narrow colour) stay flat. Out-of-row like `Signed360`; the in-row refinement budget remains the turbovec nibble