0xC0_RELOG0 contained non-hex digits (R, L, G) producing a SyntaxError
at module import, which made core pulse unable to load the GloVe-backed
manifold. Replace the wordplay constant with the equivalent integer
literal from the comment (3236855408) so the deterministic seed is
preserved and the import path is restored.
Audit of the one-mutation-path invariant (ADR-0021 §3) found three leaks
where pack authority or session-state writes could substitute for coherence
judgment. All three landed fixes or partial closures in this push.
Leaks closed:
- Leak A: pack vocab defaulted to COHERENT — flipped to SPECULATIVE in
language_packs/{compiler,schema}.py; docstring corrected to align with
ADR-0021 (it was rationalizing the leak).
- Leak B: vault.recall was epistemic-blind — VaultStore.store() now stamps
every entry with EpistemicStatus (default SPECULATIVE); recall(min_status=)
filters to admissible-as-evidence tier. All 4 vault-write sites updated.
- Leak C (write-side): generate/proposition.py:198 stored articulated
propositions unmarked — now stamps SPECULATIVE, breaking the
fabrication-feedback loop in principle. Read-side audit of 5 call sites
is the residual.
New architectural invariants (tests/test_architectural_invariants.py):
- INV-21: one-mutation-path allowlist (caught Leak C on first run)
- INV-22: pack lexicon default is SPECULATIVE (Leak A guard)
- INV-23: vault recall epistemic-aware (Leak B guard)
New eval lanes:
- teaching_injection_resistance — ships GREEN at 1.00/1.00/0 (the
structural anti-injection claim is real and measurable)
- refusal_calibration — honest gap: 0% refusal, 0% fabrication
- contradiction_detection — honest gap: 50% flag via versor-delta heuristic,
100% false-positive; motivates the proper coherence-checker
- articulation_of_status — honest gap: 0% speculative articulation, 60%
false certainty; output-side leak surface
New benchmarks:
- benchmarks/footprint.py — total deployed runtime is 7.06 MiB
(109,358x smaller than Llama 3.1 405B, runs offline, no GPU)
- benchmarks/learning_curve.py — monotonic + replay-deterministic curve
per lane
Documentation:
- docs/truth_seeking_schema.md — foundational architectural commitment,
five rules, mapped to human failure modes, leaks published openly
- evals/CLAIMS.md — five-tier public claims doc; Tier 4.5 publishes
known gaps with named fixes; verification contract at top
- README.md — new pillar between algebraic substrate and language pillar
Includes in-flight formation pipeline scaffolding (formation/, tests/formation/,
docs/formation_pipeline_plan.md) and minor CLI/contracts/gitignore edits
that were already in the working tree at session start.
Verification: 798 passed, 2 skipped, 1 deselected (pre-existing pack-count
test drift unrelated to schema changes).
ADR-0021 v1 schema land. epistemic_status is a position in the revision
graph, not a source-trust tier — coherence is the only admission signal.
Surfaces:
- teaching/epistemic.py: EpistemicStatus enum (COHERENT, CONTESTED,
SPECULATIVE, FALSIFIED); ADMISSIBLE_AS_EVIDENCE = {COHERENT}.
- PackMutationProposal.epistemic_status (default SPECULATIVE) + immutable
with_status() updater.
- ReviewedTeachingExample.epistemic_status (default SPECULATIVE);
orthogonal to acceptance per ADR §Schema impact.
- LexicalEntry.epistemic_status (default "coherent" for seed; absent in
JSONL is treated as the seed default — no retroactive tagging).
- compute_trace_hash + trace_hash_from_result + pipeline.py fold the
load-bearing proposal's epistemic_status into the trace hash so
replay detects different epistemic frames.
Non-hardening invariant (ADR-0021 §2): tests/test_epistemic_invariants.py
asserts no final/frozen/axiom/permanent flag on PackMutationProposal or
ReviewedTeachingExample, and EpistemicStatus contains no source-trust
tier names.
Docs: docs/runtime_contracts.md gains an Epistemic surface section.
Lanes green: smoke 27/27, teaching 10/10, packs 6/6, runtime 19/19,
cognition eval 100%.
Adds 15 lexical entries (071-085) extending the cognition pack with
rhetoric, metaphor, narrative, and writing-style vocabulary. Layer 1
of the work plan recorded in evals/compositionality/gaps.md and
evals/cross_domain_transfer/gaps.md: lexical scaffolding only, no
new operators. Building first-class metaphor / narrative / style
support remains correctly downstream of the cross-domain-transfer
literal case working (now closed in commit 57a6174).
New entries:
071 metaphor 076 voice 081 figure
072 simile 077 style 082 symbol
073 analogy 078 register 083 image
074 narrative 079 tone 084 discourse
075 story 080 rhetoric 085 account
Each entry follows the existing pack convention: NOUN pos, four
semantic_domains, morphology_tags=["noun"], seed provenance. The
domains anchor on rhetoric.*, language.figure/discourse/style,
cognition.*, and meaning.* clusters that integrate with the
existing pack vocabulary.
Pack-level updates:
- manifest.json checksum recomputed against the bytes actually
written to disk (per CLAUDE.md Semantic Pack Discipline).
- version bump 1.1.0 -> 1.2.0.
- test_core_semantic_seed_pack.py last-entry assertion updated
from 070 to 085.
Verification: probe "What is X?" against the new vocabulary grounds
cleanly in the pipeline (narrative 7 hits, style 9, rhetoric 8,
analogy 9 vault matches; metaphor produces a coherent surface
despite zero vault hits, consistent with the field-geometry
characterisation in the adversarial-identity calibration probe).
CLI suites packs / smoke / cognition / teaching / runtime all pass;
no regression.
What this does NOT do (deferred by design):
- No metaphor / simile / narrative operator at the proposition-
graph layer. ADR-0018 forbids building operators ahead of
eval evidence; these become a Phase 3 v3 (or Phase 4) candidate
once cross-domain transfer with selectivity has its own eval
lane.
- No first-class is_like(A,B) relation distinct from is(A,B).
Same reasoning - downstream of compositionality engineering.
- No persona/style work on the output side. That belongs in
persona/motor.py per the cross_domain_transfer/gaps.md
architectural sketch.
The entries serve as substrate for future eval lanes that probe
these capabilities specifically (metaphor-comprehension,
narrative-coherence, register-control). When those lanes are
authored, the vocabulary needed for the probes is already grounded.
- grammatical-coverage holdout v1: 52 cases across all 13 constructions, 100% pass
- zero-code-domain-acquisition lane: contract + 3 surprise domains (kinship,
calendar, color) with vocabulary, relations, axioms, teaching examples,
and dev prompts; pack closure verified for all three domains
- he_core_cognition_v1: 20 entries in Hebrew script with morphology decomposition
(triliteral roots, binyanim, aspect/person/gender/number); depth_root role
with fail_closed OOV policy
- grc_logos_cognition_v1: 20 entries in polytonic Greek with morphology
decomposition (stems, prefix/suffix chains, declension class, tense/voice/
mood/person); depth_relation role with fail_closed OOV policy
Implements the English Supervised Seeding Epoch (V1):
- language_packs/en_seeder.py: downloads GloVe-6B-50d, projects each
token embedding through a CGA lift into Cl(4,1) via construction_seed_versor,
validates the versor invariant, and registers the word in VocabManifold.
- scripts/run_pulse.py: replaces the mock 10-word hash vault with the
live VocabManifold. Injection now uses TextProjectionHead.project()
against the seeded vocab; vault_recall queries VocabManifold.nearest().
Hash fallback retained for words absent from GloVe (OOV tagged fallback).
The CGA lift preserves semantic neighbourhood: words close in GloVe
cosine space map to versors that are geometrically proximate in Cl(4,1)
inner product space, so nearest() returns semantically coherent results
rather than hash-proximity artefacts."
Add RuntimeConfig with English default output policy, wire output language through runtime/frame selection/generation/CLI, preserve language metadata in mounted manifolds, and add runtime/CLI policy tests.
* Make core CLI help robust and intuitive
* Package runtime support modules for core CLI
* Add CLI help and doctor tests
* Fix CLI trace help and pack listing
* Export language pack listing helper
* Bootstrap repo root for console runtime imports
* Align trace formatter with Proposition schema
* Cover real trace payload formatting
Compile language-pack features into even-grade rotors, apply canonicalization after alignment nudges. Preserves holonomy parity across token counts. 231 tests passing.
Root was 0.17, stem 0.10 — not heavy enough to anchor same-root forms
against the diverging suffix rotor (ים landing on a different axis).
New hierarchy:
triliteral root 0.13 → 0.22 (shared identity, strongest anchor)
root 0.17 → 0.30 (primary root geometry)
stem 0.10 → 0.18 (secondary, same-root forms cluster here)
inflection role 0.02 → 0.015 (key label, minimal)
inflection value 0.05 → 0.03 (number/gender/etc, perturbation only)
prefix 0.05 → 0.03 (per-position decay preserved)
suffix 0.04 → 0.02 (per-position decay preserved)
This ensures דבר and דברים both orbit the D-B-R root point closely
enough that cga_inner(singular, plural) > cga_inner(singular, unrelated),
while still encoding morphological distinctions as measurable offsets.
- VocabManifold.update(): replace a stored versor in-place (grade-norm
enforced, KeyError if word missing)
- compile_entries_to_manifold() now also returns entry_id->surface map
as a second return value; callers that only need the manifold use [0]
- _alignment_nudge_rotor(): slerp-style rotor that rotates source toward
target by edge.weight * ALIGNMENT_NUDGE_STRENGTH (0.06) — small enough
to preserve intra-pack geometry, large enough to pull cross-lang pairs
into proximity
- load_pack(): after both manifolds are built, loads alignment.jsonl,
resolves entry_id->surface on each side, applies nudge to source versor
using the already-compiled target versor as the attractor
- _is_hebrew_root() guard retained; _triliteral_root() Hebrew-only
- Guard _triliteral_root() rotor to Hebrew-script roots only; Greek/other
scripts now use the root: rotor alone (0.17) — no spurious uppercase
Unicode blade collision from the romanization fallback path
- test_structured_morphology_improves_same_root_hebrew_resonance: replace
tag_only baseline (now empty since legacy tags stripped) with no_morphology
baseline (domain+pos+lemma+surface only), which is the honest comparison
for what structured morphology contributes post-migration
- Strip root: tags from all 7 grc_logos_micro_v1 lexicon entries
- Refresh grc_logos_micro_v1 manifest checksum
- Remove _uses_legacy_root_tags() and _apply_morphology_tags() from compiler
- _entry_to_coordinate() now has a single morphology path: structured when
morphology_id resolves, tag fallback only for packs with no morphology data
- load_pack() registry loading made unconditional for packs with morphology.jsonl
These were the pre-rename originals superseded by he_logos_micro_v1 and
grc_logos_micro_v1. Their manifests have checksum mismatches (visible via
`python -m language_packs list`) because the lexicon content was updated
without re-checksumming before the rename.
No tests reference these pack IDs. No code loads them. Leaving broken
packs in data/ creates false noise in `list` output and risks a future
agent loading the wrong pack by prefix match.
Three subcommands:
compile <pack_id> — compile a pack, verify checksum, print manifold stats
verify <pack_id> — checksum-only verification without full compilation
list — list all available packs with role and entry count
Checksum rule (AGENTS.md) is enforced by design: the CLI reads bytes from
disk via read_bytes() as the single source of truth, never from in-memory
strings. This makes the compiler an auditable D0 instrument — not just an
importable function.
Exit codes: 0 = success, 1 = checksum mismatch or pack not found.
Permanent lesson: checksums must be computed from bytes-on-disk after
git writes the file (unicode-escaped JSON lines), not from in-memory
Python strings before serialization. The compiler CLI must do:
checksum = hashlib.sha256(Path(lexicon_path).read_bytes()).hexdigest()
after writing the file, not before.
he_logos_micro_v1: ea1ac85d...
grc_logos_micro_v1: 1fea9d9c...
Root cause: all 7 Logos words shared 'logos.core' as first domain, making
them cluster into a single blob. The misaligned triple (word ↔ ראשית,
word ↔ πνεῦμα, דבר ↔ ἀρχή) scored 0.379 vs aligned 0.299 because
cross-language 'spirit' and 'beginning' were geometrically indistinguishable
from 'word' when logos.core dominated the rotor composition.
Fix: each Logos word now has a UNIQUE leaf domain as its FIRST (highest-
weight) domain that it shares ONLY with its cross-language counterparts:
word/דבר/λόγος → logos.utterance.word
beginning/ראשית/ἀρχή → logos.genesis.origin
light/אור/φῶς → logos.illumination.photon
life/חיים/ζωή → logos.vitality.animate
truth/אמת/ἀλήθεια → logos.aletheia.verity
spirit/רוח/πνεῦμα → logos.pneuma.breath
create/ברא/κτίζω → logos.ktizo.formation
logos.core retained as second domain (background cohesion, lower weight).
divine.revelation removed from spirit/רוח/πνεῦμא (was polluting alignment).
Packs renamed: he_logos_v1 → he_logos_micro_v1, grc_logos_v1 → grc_logos_micro_v1
(test expects *_micro_* IDs).
All three manifests updated with correct D0 checksums.
- manifest checksum updated to match actual bytes-on-disk
(eca36a3d... — line endings differ between Python serialization and git)
- probe/repl.py: add sys.path.insert so it resolves language_packs
from repo root when run directly
