diff --git a/.claude/board/EPIPHANIES.md b/.claude/board/EPIPHANIES.md
index 7128b8e..17f7de4 100644
--- a/.claude/board/EPIPHANIES.md
+++ b/.claude/board/EPIPHANIES.md
@@ -15,6 +15,151 @@
 
 ## Entries (newest first)
 
+## 2026-06-23 — Live 2026 receipt for the semantic-compiler thesis: bardioc is actively extending OGIT's Auth symbol table with a linker-phase external-IAM bridge (probably Zitadel)
+**Status:** FINDING (shape-grounded; external system not named in-file → [H], not [G])
+**Scope:** addendum to the 2026-06-22 "OGIT was already a semantic compiler's symbol table" entry below × Auth-domain dating × the AuthStore-mapping pattern × the queued 0x0BXX cross-walk
+
+The 2026-06-22 entry below argued from the OGIT *shape* that bardioc
+built a semantic compiler. This is a **dated receipt** that they are
+STILL treating OGIT as the canonical symbol table — and that the
+current extension is a textbook **linker / name-resolution** phase.
+
+What the `NTO/Auth/` dates show ([G], read from `dcterms:valid`):
+
+- **The IAM core is arago's own, from 2018** — `Organization`,
+  `OrgDomain`, `Account`, `Application`, `Role`, `RoleAssignment`,
+  `Team`, `DataScope`, all `start=2018-01-01`, creator "arago GmbH".
+  This **predates Zitadel's prominence** (open-sourced ~2020–2022), so
+  the resource model is convergent-universal-IAM, NOT copied from
+  Zitadel.
+- **A January 2026 batch by `Pablo Perez`** adds foreign-key-shaped ID
+  attributes — `organizationId`, `accountId`, `applicationId`,
+  `scopeId`, `configurationData` (all `start=2026-01-12`) — plus the
+  `ApplicationContent` entity (`2026-01-14`). They hang off the
+  `Configuration` entity, described as "individual configuration for an
+  organization, user, application or scope **registered in hiro
+  knowledge core**", `belongs Organization`.
+
+The tell ([H] — pattern, not a named string): **adding FK ID columns
+is what you do to bridge to an EXTERNAL system keyed by those
+identities.** You don't add `organizationId`/`applicationId` columns to
+your OWN native entities — you already have typed edges. You add them to
+point at someone else's primary keys. The config blob lives in HIRO,
+keyed by the external IAM's org/app/user/scope IDs. That's the graph
+*side* of a bridge; the IAM lives elsewhere. Zitadel is the most likely
+external system (its `org_id / project_id / app_id / user_id` are
+exactly these four FK shapes; matches the operator's stated stack) but
+**no file names Zitadel** — hence [H].
+
+Caveat that keeps it honest: OGIT's Auth domain ALSO carries a
+**Zanzibar-relation shape** — `edgeRule` / `vertexRule` attributes
+(2018) + membership verbs (`isMemberOf`, `assigns`, `assumes`,
+`belongs`, `consents`, `uses`). So the domain is positioned to host
+both a Zitadel-resource binding AND an Ory/Keto relation-tuple binding
+— exactly the operator's earlier-this-session framing ("zitadel,
+zanzibar, ory/keto become preminted class profiles").
+
+Two consequences:
+
+1. **Strengthens the semantic-compiler thesis with a fresh receipt.**
+   The 2026-06-22 entry inferred compiler-grade discipline from a static
+   read. This shows the discipline is *live*: in 2026 they extend the
+   symbol table with external-symbol resolution — the linker phase of
+   `OGAR-AS-IR §1`, actively in use. Not a fossil; a running compiler.
+
+2. **The `Configuration`-keyed-by-external-IDs entity IS the OGIT-side
+   precedent for the "AuthStore class that does the mapping"** the
+   operator specified earlier this session, and informs the queued
+   `0x0BXX` auth-domain cross-walk (`OGIT-DOMAIN-LIFT-CATALOGUE.md` Auth
+   row). bardioc already built the bridge node; OGAR's job is to give it
+   a classid and resolve Zitadel/Zanzibar/Keto as preminted profiles.
+
+Evidence: `vocab/imports/ogit/NTO/Auth/attributes/{organizationId,
+accountId,applicationId,scopeId,configurationData}.ttl` (all
+`2026-01-12`, Pablo Perez); `entities/Configuration.ttl` (2018 class,
+2026 attribute list); `entities/ApplicationContent.ttl` (`2026-01-14`).
+Cross-ref the entry below + `docs/OGAR-AS-IR.md` (linker phase).
+
+## 2026-06-22 — OGIT was already a semantic compiler's symbol table — bardioc built the structural half deliberately, externalized behaviour to HIRO, never unified the two halves
+**Status:** FINDING (shape-inference from the OGIT artifact, not insider history)
+**Scope:** OGAR-AS-IR provenance × the structural/behavioural-arm split × what OGAR's actual contribution is
+
+Question posed: from the shape of OGIT, how likely is it that bardioc
+(arago's HIRO/Bardioc engine, the original OGIT authors) discovered the
+"semantic compiler" superpowers OGAR articulates?
+
+Assessment, reasoning purely from the OGIT artifact read end-to-end
+(NTO 72 domains + SGO upper ontology + MARS XSD + `extract_classes.py`),
+not from any insider knowledge:
+
+**High likelihood they discovered and EXPLOITED it operationally; low
+likelihood they FRAMED it as a compiler.** They built a thing that IS a
+semantic compiler and described it in Semantic-Web vocabulary
+(`rdfs:`/`owl:`/`dcterms:`), not compiler-engineering vocabulary.
+
+The discipline in the artifact is the tell — these are [G] (visible in
+the files), not inference:
+
+- **Symbol table with typed signatures** — SGO's 176 verbs, separately
+  versioned, each with `ogit:from-to` domain→range typing.
+- **Type system with closed constraints** — `validation-type "fixed"` +
+  exhaustive `validation-parameter` enums (round-trippable).
+- **Structural typing with cardinality** — `mandatory-/optional-/indexed-attributes`.
+- **Capability/interface declaration** — `ogit:allowed ([verb target])`.
+- **Module/namespace layering** — `ogit:scope "NTO"`/`"SGO"`; NTO/SGO/SDF split.
+- **Explicit dependency DAG** — MARS A→R→S→M `dependsOn` chain.
+- **Codegen back-end** — `extract_classes.py` lowers XSD/OGIT → rendered tables.
+- **IR-as-canonical-source** — OGIT was the source; HIRO consumed it;
+  automations were driven FROM the ontology.
+
+Most RDF ontologies are loose, under-typed, aspirational. OGIT is none
+of those. `validation-type "fixed"` with exhaustive parameter lists AND
+a Python extractor that preserves them is compiler-grade thinking wearing
+Semantic-Web labels.
+
+**The sharpest single piece of evidence:** OGIT carries ONLY the
+structural arm; the behaviour lived in HIRO (Elixir `gen_statem`,
+automation rules — `ELIXIR-HIRO-PREFETCH.md`). That separation —
+declarative schema here, runtime behaviour there — IS the
+structural-arm / behavioural-arm split this workspace "rediscovered."
+bardioc had it years ago.
+
+On OGAR-AS-IR's own six IR-shape tests, OGIT satisfies ~3 of 6 by
+construction: typed-signature (yes), IR-is-canonical (yes), named-lowering
+(partial — `extract_classes` is one, unlabeled); but effect-annotations
+(no — effects lived in HIRO, not OGIT), SSA (no), semantic-preservation
+guarantee (no explicit one). That profile is precisely "a disciplined
+STRUCTURAL IR with the behavioural half externalized."
+
+**What they did NOT do — and what OGAR's actual contribution is:** the
+UNIFICATION. "These are two arms of ONE IR; the structural arm lowers to
+N back-ends; the behavioural arm stays in the Core; the same address
+resolves both." bardioc had two systems (OGIT + HIRO) with a "HIRO reads
+OGIT" seam, not one IR with two arms. OGAR is not discovering the
+superpower — it is RENAMING what bardioc built (in compiler vocabulary)
+and UNIFYING the two halves they kept apart.
+
+Consequence for how we talk about OGAR: the `OGAR-AS-IR` line "the docs
+were already compiler-shaped, just not labeled" applies one level down
+to OGIT itself. Honest framing in any external-facing material: OGAR
+stands on a deliberately-engineered semantic-compiler symbol table
+(OGIT) and contributes the IR unification + the compiler-vocabulary
+framing, NOT the underlying discovery. Crediting bardioc's structural
+discipline is both accurate and strengthens the claim (the substrate is
+battle-tested, not speculative).
+
+Fences (this is shape-inference, grade honestly):
+- "[G] the shape exhibits compiler properties" — strong, evidenced in files.
+- "[H] bardioc consciously knew they were building a compiler" — inference
+  from discipline; plausible but unprovable from the artifact alone.
+- "[S] they had the full IR-discipline OGAR articulates" — no; the 6-test
+  profile (3/6) falsifies this. The unification is genuinely OGAR's.
+
+Cross-ref: `docs/OGAR-AS-IR.md` (the framing), `docs/HIRO-IN-CLASSES.md`
+(the bardioc-efficiency story), `docs/ELIXIR-HIRO-PREFETCH.md` (HIRO =
+the behavioural arm), `docs/MARS-TRANSCODING.md` (the XSD calibration that
+exercised the structural arm).
+
 ## 2026-06-22 — The "latent re-vendor bug" was a false premise; exports/ is a STAGING tier, not a permanent home (operator-decided)
 **Status:** FINDING
 **Scope:** vocab/ tree model × verify-before-acting × correcting a prior session's claim