cbfc8ccbf7
The CORE-Logos "Crown Proof" — ADR-0015's claim that cross-language holonomy
closure is THE validation gate of meaning — was measured against a criterion
registered before the run, and refused.
G1 separation, all negatives AUC 0.557 (chance 0.500, bar 0.80) FAIL
G2 hardest class, cross-pairing AUC 0.664 (bar 0.75) FAIL
G3 word-order sensitivity 0.644 (bar 0.90) FAIL
G4 no collapse re-entry 0 lost (bar: 0) PASS
G4 is the half that succeeded: the repairs work. The claim is what failed --
which is the outcome the registration called "worth as much as a GO", because
it retires the largest piece of unearned design in the system.
Both preconditions had to be BUILT before the question could be asked at all,
which is why the 2026-06-14 negative constrained nothing: the shipped compiler
collapsed 37-53 coordinates, and holonomy_encode never closed. Measured on the
repaired ground with a genuine loop, over 1,016 aligned pairs and 58,375
mechanically-generated negatives in three classes, nothing hand-picked.
Diagnosis, not just a number: the encoding reacts more strongly to REORDERING a
clause (median deviation 41.2) than to CHANGING WHAT IT IS ABOUT (32.2). It
measures path shape, not content -- and that follows from the algebra, since
permutation acts on an ordered product through non-commutativity, a first-order
effect, while substituting one of three factors perturbs it only through that
factor. The negative classes do order correctly by meaning-distance
(aligned 27.2 < lexical 32.2 < order 41.2 < cross-pair 53.2), so the geometry is
weak signal rather than noise -- a correlation where the design asked for a gate.
Consequences, all landed here:
* ADR-0005 and ADR-0015 AMENDED to record the refusal. The three-language
architecture, the pack contract, morphology-as-structure and Hebrew root
folding into geometry all survive untouched; only the claim that alignment
or closure VALIDATES meaning is retired.
* tests/test_fa1_gate_verdict.py is the tripwire: if a future encoding makes
the gate real, it fails, and the amendment is withdrawn in the same commit
that replaces it with a proof. Two sabotages observed red.
* allow_cross_language_recall RULED by measurement (work-order item 3): it
gates vault-recall depth over field states, has no language argument and no
cross-language path, and reaches only walk_surface, which runtime_contracts
declares telemetry. Stays ON, reclassified DEPLOYMENT, flip condition
recorded. G-25's own claim that it was "the pillar's own switch" is
CORRECTED -- that read the flag's name instead of its call graph, which is
the error the audit exists to find.
* A sixth defect found while building the harness: en_collapse_anchors_v1
declares role "collapse_anchor", which is not a LanguageRole member, so
load_pack() raises. It is registered in the resolver, consumed by opening
lexicon.jsonl on a raw path that bypasses the loader, and is the target of
all 24 alignment edges that still resolve to nothing. The enum's own comment
records the identical failure once before (ADR-0097, domain_seed).
The repairs cross to the keel on their own merit regardless of the verdict:
geodesic blending removes every coordinate collision, and mount-wide edge
resolution connects 39 of the 63 dead edges. The collapse was damaging English's
own vocabulary; seeing that required no theory of meaning.
FA-1 is CLOSED: all four work-order items discharged, L2 verdict recorded as
defective-but-repairable with its central design claim retired.
[Verification]: uv run core test --suite smoke -q -> 786 passed in 239.45s, EXIT=0
(+4 tests, +15s vs the prior 782/224s -- the FA-1 pin's stated cost).
Doc-parsing pins re-run after the ADR and register amendments:
test_adr_status_governance + test_adr_index -> 321 passed;
test_flag_register -> 6 passed.
185 lines
7.6 KiB
Markdown
185 lines
7.6 KiB
Markdown
# ADR-0005: Language Pack Contract
|
||
|
||
**Date:** 2026-05-12
|
||
**Status:** Accepted
|
||
|
||
## Context
|
||
|
||
CORE operates on three foundational languages: English (default articulation
|
||
base), Hebrew (depth pack 1), and Koine Greek (depth pack 2). These are not
|
||
equivalent choices from a menu — they are the three specifically chosen
|
||
because of the range of depth they bring and because of the theological and
|
||
linguistic insight encoded in John 1:1–2, where the universe itself was spoken
|
||
into existence and John articulated that reality in both Hebrew and Greek.
|
||
This is why and how CORE finds its truth and power in design and
|
||
communication — being primary, being a core.
|
||
|
||
A loose plugin API is insufficient for these packs. The system needs a strict
|
||
contract that specifies what a language pack must provide, what invariants it
|
||
must satisfy, and what gates it must pass before activation.
|
||
|
||
## Decision
|
||
|
||
Adopt a **Language Pack Contract** as the canonical template for all CORE
|
||
language packs.
|
||
|
||
Each pack is a deterministic, versioned manifold bundle that defines:
|
||
|
||
- identity and normalization policy
|
||
- lexical inventory (lemma-first, not token-first)
|
||
- morphology
|
||
- syntax and frame templates
|
||
- semantic lift rules into CORE-native pressure
|
||
- readback rules from field state to surface language
|
||
- validation probes and alignment gates
|
||
|
||
The three canonical packs are:
|
||
|
||
| ID | Name | Role |
|
||
|-----|--------------------|-----------------------------|
|
||
| `en`| English | Default articulation base |
|
||
| `he`| Hebrew | Depth language pack 1 |
|
||
| `el`| Koine Greek | Depth language pack 2 |
|
||
|
||
Each pack must implement the same top-level contract and must target shared
|
||
field primitives. No pack may define private or opaque semantics that bypass
|
||
the shared field.
|
||
|
||
## Required Structure
|
||
|
||
```
|
||
packs/
|
||
common/
|
||
contracts/
|
||
language-pack.md
|
||
schema/
|
||
lemma.schema.json
|
||
morphology.schema.json
|
||
frame.schema.json
|
||
sense.schema.json
|
||
probe.schema.json
|
||
anchors/
|
||
trilingual-anchor-template.json
|
||
en/
|
||
pack.toml
|
||
orthography.yaml
|
||
lemmas.jsonl
|
||
morphology.jsonl
|
||
frames.jsonl
|
||
senses.jsonl
|
||
lift_rules.py
|
||
readback_rules.py
|
||
validators.py
|
||
probes/
|
||
corpora/
|
||
manifest.yaml
|
||
he/ (same structure)
|
||
el/ (same structure)
|
||
```
|
||
|
||
## Validation Gates
|
||
|
||
A pack is not considered active because files exist. A pack becomes active
|
||
only when it passes the following gates in order:
|
||
|
||
1. **Schema gate** — all data files validate against their JSON schemas
|
||
2. **Lexical gate** — lemma inventory is non-empty and internally consistent
|
||
3. **Morphology gate** — all surface forms trace to a known lemma
|
||
4. **Lift gate** — lift rules produce valid `CandidateGeometricPressure` for seed inputs
|
||
5. **Readback gate** — readback rules produce grammatical surface from seed field state
|
||
6. **Determinism gate** — all D0/D1 paths produce identical output on repeated runs
|
||
7. **Alignment gate** — pack publishes anchor records verifiable against trilingual template
|
||
8. **Coverage gate** — probe set covers minimum lexical and morphological surface
|
||
|
||
> **Amendment · 2026-07-28 — what gate 7 actually verified, measured.** The alignment gate
|
||
> presupposes that publishing anchor records establishes cross-language alignment. Measured at
|
||
> `472fc0a8`, the mechanism behind those records **destroyed** what it claimed to align:
|
||
> `packs.compiler._blend_feature_versors` ignored its strength argument and returned the target
|
||
> verbatim, so an "aligned" Hebrew token was *overwritten* by its Greek or English partner
|
||
> rather than drawn toward it, and mounting the depth packs removed 37 distinct coordinates
|
||
> (`docs/analysis/logos-substrate-collapse-2026-07-28.md`). The eight gates are also not
|
||
> enforced as a sequence anywhere in the loader — `gate_engaged` is a single manifest boolean
|
||
> consulted by `sensorium/registry.py`, and the seven other gates have no code that runs them.
|
||
>
|
||
> Two things follow, and only two. **(a)** Alignment must be *interpolation on the versor
|
||
> group*, never replacement — `algebra/rotor.py` has shipped that operation since the algebra
|
||
> layer's first commit. **(b)** Gate 7's stronger sibling — that alignment establishes
|
||
> resonance strong enough to *validate meaning* (ADR-0015's Crown Proof) — was measured and
|
||
> refused: `docs/analysis/fa1-holonomy-gate-verdict-2026-07-28.md`. The alignment data itself
|
||
> stands: 24 aligned concepts across three languages, and Hebrew root folding into geometry is
|
||
> unaffected. What is retired is the claim that alignment, by existing, proves anything.
|
||
|
||
## Design Rules
|
||
|
||
### 1. Lemma-first
|
||
Tokens are surface realizations. Lemmas are the stable lexical unit. The
|
||
ingest and field-state layers operate on lemmas, not tokens.
|
||
|
||
### 2. Deterministic normalization
|
||
Normalization must be deterministic and script-aware. Unicode normal form NFC
|
||
is the minimum. Script-specific rules (right-to-left, vowel pointing, accent
|
||
diacritics) are defined per pack in `orthography.yaml`.
|
||
|
||
### 3. Native morphology
|
||
Morphology is part of meaning-bearing structure, not optional metadata.
|
||
Hebrew stems (Qal, Niphal, Piel…) and Koine Greek aspects (aorist, perfect…)
|
||
are semantically load-bearing. Morphology is a first-class pack surface.
|
||
|
||
### 4. Shared field target
|
||
Semantic lift must target shared CORE field primitives. No pack-private
|
||
semantic space is permitted. What cannot be expressed in shared primitives
|
||
must be proposed as a new shared primitive, not hidden inside a pack.
|
||
|
||
### 5. Readback is owned locally
|
||
Each pack owns its articulation rules and ambiguity management. Readback is
|
||
not a global postprocessor — it is part of the pack contract.
|
||
|
||
### 6. Alignment is explicit
|
||
Packs must publish anchor records that are verifiable against the
|
||
trilingual anchor template. Cross-language coherence is a probe surface,
|
||
not an emergent hope.
|
||
|
||
## Runtime Boundary Honesty
|
||
|
||
Files that define a semantic boundary not yet fully specified must raise
|
||
`NotImplementedError` at exactly that boundary. This is not fake code — it
|
||
is an honest contract: the structure is real, the interface is real, the
|
||
unimplemented edge is visible and loud. Nothing is silently skipped.
|
||
|
||
## Rationale
|
||
|
||
This serves the three engineering pillars:
|
||
|
||
- **Mechanical Sympathy** — a small, explicit, cheap-to-validate pack surface
|
||
that integrates directly with the field-state model.
|
||
- **Semantic Rigor** — all packs target shared field primitives instead of
|
||
inventing language-private meanings. Exactness is non-negotiable.
|
||
- **Third Door** — standard token-plugin patterns were considered and
|
||
rejected. This contract is designed for CORE's geometry specifically.
|
||
|
||
## Consequences
|
||
|
||
**Easier:**
|
||
- Every future language pack follows the same geometry and QA path.
|
||
- English, Hebrew, and Koine Greek can be aligned by shared anchors and
|
||
deterministic probes.
|
||
- Gate failures are explicit and traceable.
|
||
|
||
**Harder:**
|
||
- Pack creation requires real morphology, lift, and readback logic.
|
||
- Shortcuts that bypass field primitives are structurally forbidden.
|
||
|
||
**Forbidden:**
|
||
- Opaque embedding tables.
|
||
- Black-box semantics.
|
||
- Pack-local meaning systems that bypass the shared field.
|
||
|
||
## Alternatives Considered
|
||
|
||
- **Loose plugin API:** Rejected. Too weak to enforce invariants.
|
||
- **Dictionary-only packs:** Rejected. Lexicons alone do not articulate or lift.
|
||
- **Tokenizer-first packs:** Rejected. Tokens are surface artifacts, not the
|
||
stable unit of design.
|
||
- **LLM-as-extraction-engine:** Rejected. Introduces a D3 nondeterministic
|
||
oracle into the normalization site. See ADR-0002 for the ingest-layer ruling
|
||
on determinism classification.
|