docs: geometric retrieval + §4 managed-memory cure (live) and autonomous superseding self-reification (design)

- 06 §2.5: shipped 2026-08-14 substrate — structure-gated geometric retrieval
  (P@5 0.700, semantic-not-lexical) and the §4 write-barrier + generational GC
  cure (store 1.616GB→38.5MB, RSS→82MB, zero loss, reboot-proven) + LLM token
  telemetry.
- 06 §4.1: autonomous, continuous, superseding self-reification on the heartbeat
  — reification as an operation OF the engram; explicit reify/rename/run-a-pass
  as the degenerate case; no gate/pause; flat + overlapping domains; contextual
  importance; supersession-as-residue; secondary-soul validation, flag-gated.
- 06 §6.0: relating as the primitive (one capability; the rest is terrain);
  perspective calculated via geometric transformations.
- 06 §6.4: reasoning as constructive self-argument governed by the verifier.
- 03: supersession-as-residue note; live geometric retrieval on route_search.

docs/architecture/03-data-and-memory.md

@@ -165,6 +165,13 @@ Engram nodes are immutable (`memory.el:64-69`). The model is:
   returns both ids so the caller re-points. This is the `supersedes_id`
   pattern: new node linked, old preserved, full audit trail.
 
+> **Supersession is residue, not garbage.** The superseded node is the *trail of
+> how the current understanding was reached* — kept deliberately, because sometimes
+> the truth was in the **old** idea even when the old idea was not itself the truth.
+> This is what lets autonomous self-reification (`06` §4.1) run ungated: every
+> rename/re-cluster supersedes into this residue chain, so nothing it does is ever
+> destructive — the safety is *after* the act, not a gate before it.
+
 > **The hole to know about.** The raw runtime `engram_forget` **does** hard-delete
 > (frees node + edges, `el_runtime.c:7647`), and the engram HTTP route
 > `DELETE /api/nodes/:id` calls it directly (`server.el:322-328`). Immutability
@@ -228,6 +235,11 @@ Retrieval is **spreading activation, not query matching**:
 cosine(query, target)` — multiplicative, top-N, with the two-layer
 background → working-memory promotion (`README.md:27-36`; `el_runtime.c:5892+,
 6094+`). `mem_recall` / `/api/activate` fire this and mutate WM; `mem_search` /
-`/api/search` are passive lexical scans. The cognitive API's `begin_session` and
-`compile_ctx` return a **bounded projection** of the activated set, never the raw
+`/api/search` are passive lexical scans — **but as of 2026-08-14 the live
+`route_search` runs structure-gated *geometric* retrieval**
+(`engram_retrieve_geometric_json`; held-out **P@5 = 0.700**, semantic not lexical —
+`skill` returns skill nodes and *rejects* the false-positive `rainfall`), with the
+old lexical scan retained at `/api/search-lexical` (see `06` §2.5). The cognitive
+API's `begin_session` and `compile_ctx` return a **bounded projection** of the
+activated set, never the raw
 graph (doc 02, §2).