Skip to content

Commit e80d8d8

Browse files
committed
docs(discovery-map): §4.1 runtime synergy receipts — lance-graph PR #477/#478
Cross-referenced PR #477 (SoA envelope + HHTL audit + three-tier model) and #478 (singleton-to-snapshot) against today's D-* findings. Result: several findings are confirmed by runtime CODE (not just summaries), one is revised, one gradient is convergent-but-unwired on both sides. Confirmations [G] — runtime code grounds the OGAR finding: - D-MORTON / D-CASCADE / D-XOR2(containment): hhtl.rs ships NiblePath{path:u64,depth:u8}, FAN_OUT=16, MAX_DEPTH=16 (16 nibbles = 64 bits), parent()=path>>4, child(n)=(path<<4)|n, is_ancestor_of = a single prefix-shift kernel, common_ancestor = LCA. The nibble address algebra is CODED runtime-side. - D-IMMAT / D-COLUMNAR / ADR-022: the three-tier-model invariant ("zero-copy creation->tombstone; nothing serialized; in-memory store unchanged") IS the immaterialized finding; emit()/ CollapseGateEmission scheduled for removal (the serialization path is being deleted). - D-COLUMNAR / D-LANCE: soa_envelope.rs SoaEnvelope trait — columns() -> &[ColumnDescriptor], repr(C) descriptors, zero-copy row_le/ column_le views, as_le_bytes(), verify_layout() gate. The Morton-keyed columnar+projection framing is the ABI line-for-line. - D-DELTA: cycle()->u32, "a Lance version IS a coherent LE envelope", last_active_cycle[u32;N] per-row write guard, DatasetVersion(v)-> (v+1). Version=frame, per-row cycle stamp=changed-cell delta. Promotes D-DELTA [H]->[G]. - D-AMORT: PR #478 test (never-mutated->const codebook keep; mutated ->SoA snapshot nudge) operationalizes the amortization gate's storage instance. Revision [warn] — D-META64: MailboxSoA has SEPARATE columns edges[CausalEdge64;N] and meta[MetaWord;N], so the meta layer is a dedicated MetaWord, NOT CausalEdge64 (the edge encoding). The "CausalEdge64 = meta" reading is contradicted by the shipped column layout. Reconcile: the 2^6/48/16 bit-budget likely applies to MetaWord. Updated the D-META64 ledger entry to H->REVISE with the correction. Convergent-but-unwired [next step] — D-MAXAMORT/D-CARD/D-PAL256 (the meta->content->spatial gradient) matches high_heel.rs's module-doc legend "HEEL=scent / HIP=palette / TWIG=SpoBase17 / LEAF=full planes", but the q4 audit says that legend is a COMMENT — "no code routes by prefix." HIP=palette is literally D-PAL256 at a cascade level. Highest- value next materialization: route high_heel.rs's commented cascade by hhtl.rs's NiblePath prefix; today's findings are the spec that joins the two files (which "communicate in module docs, not in code"). Related structure: the three-tier model (data/lifecycle/supervision) is orthogonal to the meta->content->spatial gradient — noted, not conflated. Kanban 6-phase = the ActionState lifecycle (ADR-001). qualia[QualiaI4_16D;N] = the Quintenzirkel qualia codebook (#470). Docs-only. PII abort-guard (word-boundary): CLEAN. cargo check: clean. https://claude.ai/code/session_01PBTGaPCSnnt6u3pjXpbLwY
1 parent 153246f commit e80d8d8

1 file changed

Lines changed: 56 additions & 1 deletion

File tree

docs/DISCOVERY-MAP.md

Lines changed: 56 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -85,7 +85,7 @@ two halves of a cell. ADR‑026 names the cascade that ties them.
8585
| D‑PAL256 | palette256 = one PQ subspace's 256 centroids (CAM = 6×256, PR #477) | G | ADR | ADR‑024 ||
8686
| D‑CAM | CAM‑PQ = 6 roles × 256 = 48‑bit fingerprint per word (`cam_codes.bin`) | G | EPIPHANY | SYN §3 | D‑PAL256 |
8787
| D‑CARD | cascade per‑axis cardinality ↔ codebook size: 256↔palette, 4096↔`MAX_VOCAB` | G | EPIPHANY | SYN §3,§7 | D‑CASCADE, D‑CAM |
88-
| D‑META64 | `CausalEdge64` = meta: `2⁶` role‑mask **+** 48‑bit CAM + 16‑bit headroom, one 64‑bit word | H | EPIPHANY | SYN §9.6 | D‑CAM, `[per rt]` CausalEdge64 |
88+
| D‑META64 | meta = `2⁶` role‑mask + 48‑bit CAM + 16‑bit headroom. **⚠ PR #477 revises:** `MailboxSoA` has **separate** `edges [CausalEdge64;N]` *and* `meta [MetaWord;N]` columns → **the meta is `MetaWord`, not `CausalEdge64`** (the edge encoding). Reconcile (§4.1). | H → REVISE | SYN §9.6 + §4.1 | D‑CAM, `[per rt]` |
8989
| D‑BITGATE | 48→64 headroom: store the seed **iff** irreducible‑beyond‑the‑address (else compute) | H | EPIPHANY | SYN §9.6 | D‑META64, D‑AMORT |
9090
| D‑RHO | ρ = 0.9973 vs cosine (arm‑discovery aerial codebook) — the empirical anchor | G | ADR | ADR‑024 | D‑PAL256 |
9191

@@ -205,6 +205,61 @@ shader tile‑contract · SoA headroom budget. (`CausalEdge64` cardinality
205205

206206
---
207207

208+
## 4.1 Runtime synergy receipts — lance‑graph PR #477 / #478
209+
210+
Cross‑reference of the runtime‑shipped contract against today's findings.
211+
Several findings are now **confirmed by runtime code** (promotes their
212+
grade/status); one is **revised**; one gradient is **convergent‑but‑unwired
213+
on both sides** (the next materialization).
214+
215+
### Confirmations — runtime code grounds the OGAR finding [G]
216+
217+
| Finding | Runtime receipt (PR #477) | Verdict |
218+
|---|---|---|
219+
| **D‑MORTON, D‑CASCADE, D‑XOR2 (containment)** | `lance-graph-contract/src/hhtl.rs`: `NiblePath { path:u64, depth:u8 }`, `FAN_OUT=16`, `MAX_DEPTH=16` (16 nibbles = 64 bits), `parent()=path>>4`, `child(n)=(path<<4)\|n`, `is_ancestor_of` = a single prefix‑shift kernel, `common_ancestor` = LCA. (q4‑hhtl‑audit) | **the runtime SHIPS the nibble address algebra** — these are CODED runtime‑side, not just EPIPHANY |
220+
| **D‑IMMAT, D‑COLUMNAR, ADR‑022** | `soa-three-tier-model.md` invariant: *"zero‑copy from creation to Lance tombstone … no bytes leave the backing store … even then the in‑memory store is unchanged, not serialized and freed."* `MailboxSoA::emit()` + `CollapseGateEmission` **scheduled for removal** | the "immaterialized, never serialized" finding IS the runtime's ONE invariant; the serialization path is being **deleted** |
221+
| **D‑COLUMNAR, D‑LANCE (column projection)** | `soa_envelope.rs`: `SoaEnvelope` trait — `columns()->&[ColumnDescriptor]`, `ColumnDescriptor{name_id,kind,elems_per_row,row_offset}` (`repr(C)`), zero‑copy `row_le`/`column_le` views, `as_le_bytes()`, `verify_layout()` gate | the "Morton‑keyed columnar + projection" framing is the `SoaEnvelope` ABI, line‑for‑line |
222+
| **D‑DELTA (delta frames = version‑diff)** | `cycle()->u32` stamp; *"a Lance version IS a coherent LE envelope"* (soa_envelope L16); `last_active_cycle [u32;N]` = per‑row same‑cycle write guard; `DatasetVersion(v)→(v+1)` | a Lance version = a frame; the per‑row cycle stamp = the changed‑cell delta → **D‑DELTA promotes [H][G]** |
223+
| **D‑AMORT (amortize‑or‑don't‑spend)** | PR #478 mechanical test: *never‑mutated → const codebook, **keep**; mutated‑after‑init → SoA snapshot, **nudge**.* Read‑only codebooks (role keys, UDFs, `simd_caps`) stay const | the gate's storage instance, **operationalized** — const codebooks = the amortized build‑once structures; SoA = per‑cycle mutable |
224+
225+
### Revision — runtime evidence corrects a finding ⚠
226+
227+
- **D‑META64.** `MailboxSoA` columns include **both** `edges [CausalEdge64;N]`
228+
**and** `meta [MetaWord;N]` as *separate* columns. So **the meta layer is a
229+
dedicated `MetaWord`, not `CausalEdge64`** (which is the causal‑edge
230+
encoding). The operator's "CausalEdge64 = meta" is contradicted by the
231+
shipped column layout. **Reconcile:** the `2⁶`/48/16 bit‑budget reasoning
232+
(D‑BITGATE) most likely applies to `MetaWord`; `CausalEdge64` is the edge.
233+
`[per runtime session]` on the `MetaWord` layout + whether `CausalEdge64`
234+
feeds it.
235+
236+
### Convergent but unwired — both sides describe it; neither built it [the next step]
237+
238+
- **D‑MAXAMORT / D‑CARD / D‑PAL256** (the meta→content→spatial gradient) ↔
239+
`high_heel.rs` module‑doc legend **"HHTL cascade mapping: HEEL=scent /
240+
HIP=palette / TWIG=SpoBase17 / LEAF=full planes."** The q4 audit's verdict:
241+
*"this is a layout legend in a comment. **No code routes by prefix.**"* So
242+
the coarse→fine cascade (HEEL→HIP→TWIG→LEAF ≈ meta→**palette**→SPO→full) is
243+
**convergent intent on both sides and wired on neither.** `HIP = palette`
244+
is literally D‑PAL256 at a cascade level. **This is the highest‑value
245+
next materialization:** route `high_heel.rs`'s commented cascade by the
246+
`hhtl.rs` `NiblePath` prefix — the two files "communicate in module docs,
247+
not in code" (audit), and today's findings are the spec that joins them.
248+
249+
### Related structure (resonant, different decomposition)
250+
251+
- **Three‑tier model** (Tier 1 `MailboxSoA` data / Tier 2 Kanban‑Rubicon
252+
lifecycle / Tier 3 ractor supervision) is a *data / lifecycle /
253+
supervision* split, **orthogonal** to the meta→content→spatial gradient —
254+
note both, don't conflate. The Kanban 6‑phase (Planning → CognitiveWork →
255+
Evaluation → Commit → Plan → Prune) = the `ActionState` lifecycle (the
256+
OGAR Rubicon binding, ADR‑001).
257+
- `qualia [QualiaI4_16D;N]` (4‑bit, 16‑D) = the Quintenzirkel qualia codebook
258+
(BindSpace dissolution, lance‑graph #470) — a codebook column alongside
259+
`edges`/`meta`; relates to D‑CARD's codebook cascade.
260+
261+
---
262+
208263
## 5. The shape graph — the topology individual ADRs lose
209264

210265
```

0 commit comments

Comments
 (0)