Per fox 2026-06-01: same treatment as #000071 — replace the
review-archaeology structure with what we SHOULD grow. Ticket goes
from 978 lines (original Anchor6 §§1-8 design log + dav1d-review §0
retrofit) to 498 lines of directive spec. **Joseph6 stays as the
first registered example grammar** per fox's note — concrete enough
that an implementer sees what a WorldDimensionGrammar looks like
end-to-end, not abstract enough to lose its load-bearing role.
What changed in shape:
Before: §0 dav1d verdict retrofit + §§1-8 archaeology of the
original Anchor6 spec being reviewed (validate seed source,
segmentation method, mapper choice — all decisions long
since made).
After: §1-13 forward spec. Goal · Axis split · Hard constraints
(all phases) · AnchorN primitive · WorldDimensionGrammar ·
Quantization mappers (with uint256-H₁ + no-SO(3) corrections
documented inline) · π*_w_object canonicalizer with
four-identity-hash record · **Joseph6 as worked example** ·
Phase 1 deliverables (9 items) · Pre-review empirical
bench preserved as §7 · Phase 2/3 deferred · Open questions
(3 remaining; 5 closed by bench, 4 by dav1d's review) ·
Cross-references · Five-step alignment · One-line review
history at the bottom.
What changed in content: nothing material. The corrected spec from
the prior §0 retrofit IS the body now. The original Anchor6 design
log is no longer inlined — git history preserves it at commit
`862662b` (pre-rewrite tip); readers who want the rejection-by-
rejection detail go to
docs/dav1d-reviews/000070-spatial-anchor-pi-w-object--2026-06-01.txt.
Critical technical corrections preserved inline (not as "what was
fixed", but as the directive answer):
- §3.1: uint256 for H₁ position (octree depth >8 entropy
preservation)
- §3.2: rename `map_rotation_so3` → `map_rotation_euler_ypr` (no
SO(3) overclaim — quantized Euler is not SO(3) coverage)
- §4: WorldObjectRecord carries all four identity hashes
(grammar_hash, axiom_pack_hash, manifest_hash, seed_hash) for
replayability
- §5: Joseph6 ships as one example grammar; future grammars
register through the same mechanism
TICKETS.md index row also rewritten in directive voice.
doc_counts tests still pass.
Per fox 2026-06-01: replace the descriptive review-archaeology
structure with what we SHOULD grow. Ticket goes from 805 lines
(original §§1-8 design log + my dav1d-review §0 retrofit) to
389 lines of clean directive spec.
What changed in shape:
Before: §0 "Dav1d review verdict" decision-table + §§1-8
archaeology of the original Joseph6 spec being rejected.
After: §1-12 forward spec. Goal at the top, hard constraints,
composite ChainRoot identity, five bridge outcomes with
typed witness shapes, privacy class vocabulary, Phase 1
deliverables, Phase 2/3/4 roadmap, retro-validation
appendix, cross-references, three remaining open
questions, five-step alignment, one-line review history
at the bottom pointing at the archive file.
What changed in content: nothing material. The corrected spec from
the prior §0 retrofit IS the body now. The original Joseph6 design
log is no longer inlined — git history preserves it at commit
fadc50a; readers who want the rejection-by-rejection detail go to
docs/dav1d-reviews/000071-world-bridge-grammar--2026-06-01.txt.
TICKETS.md index row also rewritten in directive voice. Shorter,
less "what was wrong" + more "what to build."
Net effect: an implementer picking this up reads a forward-looking
ticket they can act on, not an archaeology of which framing was
rejected. The dav1d review history is one line at the bottom, not
the structural frame.
doc_counts tests still pass.
The 1026-line de-novo response arrives after #000070 was corrected
from fixed Anchor6 to generic AnchorN/WorldDimensionGrammar. The
bridge concept is valid and necessary; the Joseph6-coupled framing
is now wrong. Same pattern as #000070's review: §§1-8 preserved as
the design log of what was reviewed; new §0 carries the corrected
spec.
Verdict matrix (12 rows in §0 decision table):
✅ GO: bridge concept (Agreement / Translation / Embassy), Phase 1
doc-only, bridge_seed@v1 deferred to Phase 2, no audit_mode,
no SQL, bridge atlas Phase 3.
❌ REWRITE: #000070-as-Joseph6-sibling → AnchorN sibling;
chain_id=governance_policy_hash → composite ChainRoot
with optional v7-W fields (null sentinels for language-
only QA chains); single event_type='bridge' → typed
schemas per kind; privacy as Phase 3+ footnote → Phase 1
vocabulary (4 classes); #000059 "already-shipped"
overclaim → proposed/structurally-aligned unless repo
confirms.
❌ REJECT: doc-only Phase 1 flipping #000013 to "kernel_in_progress"
— correct status is "bridge_grammar_specified"; kernel_
in_progress is for actual kernel landings.
Five bridge outcomes (was three):
Agreement — grammars match on invariant set
Translation — hash-pinned adapter proves the mapping
Embassy — foreign object hosted with limited rights
Quarantine — NEW: bridge attempted, invariant validation FAILED,
record the rejection so future attempts see what broke
No-bridge — NEW: grammars genuinely incompatible, explicit
declaration that no bridge will exist
Composite ChainRoot identity:
chain_id = SHA256(canonical({
chain_id_version, history_root, governance_policy_hash,
canonicalization_version, schema_version, chunking_version,
world_manifest_hash, world_dimension_grammar_hash,
axiom_pack_hash, optional_verifier_policy_hash
}))
Language-only chains: three v7-W fields are null sentinels →
chain_id collapses to existing identity. v7-W chains: all mandatory.
Phase 1 deliverable (corrected):
1. Substrate-paper extension §"World-bridge grammar" (5 outcomes
with typed witness shapes + composite ChainRoot + privacy
vocabulary + sovereignty rule + #000070 cross-ref + retro-
validation appendix with corrected shipped/proposed framing)
2. Optional namespace stub arborist/world/bridge/__init__.py
(STATUS = "namespace_reserved")
3. NO kernels / NO bridge_seed canonicalization / NO SQL / NO
cache-key dimension / NO new audit_mode / NO verifier change /
NO KATs
4. #000013 status bumps to "bridge_grammar_specified", NOT
"kernel_in_progress"
Sovereignty rule survives unchanged (chain law local, bridge law
treaty-only, no bridge overwrites native chain law — guard against
forced-unity failure mode). Hash-pinned translators survive
unchanged. Topic-named under arborist/world/bridge/ unchanged.
Full review archived at:
docs/dav1d-reviews/000071-world-bridge-grammar--2026-06-01.txt
Folds dav1d's 1904-line review verdict into the ticket as §0 (new),
preserves §§1-8 as the design log of the original Anchor6 proposal
that was reviewed. Status flips from "open · awaiting dav1d review"
to "open · dav1d GO with rewrite · spec revision pending before any
kernel ships."
Verdict matrix:
✅ GO: generic AnchorN substrate primitive (not fixed Anchor6)
✅ GO: Joseph6 as the first registered default grammar (not THE
ontology)
✅ GO: deterministic object-state canonicalization
✅ GO: axiom/physics-loaded WorldDimensionGrammar as the scalable
path
❌ NO-GO: hard-coding H₁..H₆ as final ontology
❌ NO-GO: runtime LLM-decided dimensionality (axioms MAY propose,
only deterministic validators may accept; grammars must
freeze via grammar_hash before proof-path use)
❌ NO-GO: framing this as a "semantic verifier warrant"
❌ NO-GO: relation/event/place/agent_trace in this ticket
❌ NO-GO: SQL persistence at Phase 1
Critical technical corrections:
- H₁ MUST use uint256, not uint64 (octree position entropy at
depth >8)
- Rotation mapper is map_rotation_euler_ypr(), NOT "SO(3)"
(review §17 — quantized Euler triple, no continuous SO(3))
- Canonical record carries grammar_hash + axiom_pack_hash +
manifest_hash + seed_hash alongside the per-dimension values
- Missing privacy.class = HARD reject (no PUBLIC fallback)
- spatial-anchor-object@v1 folds into canonicalization_version
only — no new audit_mode token, no verifier_policy_hash bump
Corrected package layout:
arborist/substrate/spatial_anchor.py (AnchorN, split_anchor_n)
arborist/world/grammar.py NEW — WorldDimensionGrammar
arborist/world/pi_star/object.py (derive_world_object_record
+ 5 mappers)
arborist/pi_star/spatial_anchor_object.py NEW — registry adapter
bench/fixtures/spatial-anchor-object/known-answer-tests.jsonl
tests/test_spatial_anchor.py
tests/test_world_dimension_grammar.py
tests/test_pi_star_spatial_anchor_object.py
Implementation NOT started — arborist/world/__init__.py STATUS still
"namespace_reserved"; only bench/spatial_anchor_validation.py
(pre-review empirical bench, commit 55b651f) on disk. Review arrives
at the right moment: no production code committed against the original
Anchor6 spec yet, so the AnchorN reframe lands in the spec before
the wrong primitive ships.
Full review archived at:
docs/dav1d-reviews/000070-spatial-anchor-pi-w-object--2026-06-01.txt
Five of ten original open questions resolved by the pre-review bench
(Q1/Q2/Q3/Q8/Q9); review answers Q4/Q5/Q6/Q7; Q10 (paper-amendment
wording) is the only remaining open question — substrate paper text
needs revision to introduce AnchorN + grammar layer instead of fixed
Anchor6.
Bench-v5 (5 themes × 3 questions × 2 paths, Hermes-3-8B) AFTER the
fold-stack lift + apply_title_boost wire-up to use _title_query_tokens
shows the SAME 12 regressions as the pre-wire-up baseline:
- 5 wrong-primary picks (Dr Who, Albert/Ahmed/Alaric the third/first,
Casa Batlló error)
- 7 STRICT→HYBRID demotes on correct primaries (Spider-Man, Ampère,
Dr Syn, Dr V64, Finnish Defence Forces, Hopewell Centre, labor
economics)
Diagnosis: apply_title_boost only reranks docs ALREADY in the
candidate set. body-only retrieval (default with multi_route=False)
never surfaces "Doctor (Doctor Who)" so no fold-aware rerank can
promote it. Legacy surfaces it via the title route. FTS5 porter
stemmer handles plurals but NOT Dr→Doctor or third→III, so the
SqliteShardCorpus.fts_title method also doesn't help here without
the folds applied at retrieval-token-gen time, not rank time.
Updated ticket with the honest assessment: 5 Path A stages across
v1/v2/v3 have now proven legacy query() doesn't decompose into a
library of helpers. Three forward paths offered for fox to decide:
A. leave default at providence, accept fold regressions (env
escape hatch ARBORIST_LEGACY_QUERY=1 already shipped)
B. flip default back to legacy, treat providence as
infrastructure for cloud-query/corpus-query only
C. keep both alive long-term — separate query2 command
Re-bench legacy vs providence_query on fold themes (accent, hyphen,
honorific, brit, numeral) after the proxy memory fix.
Result (15 question-pairs through Hermes-3-8B): 12 regressions,
2 improvements, 1 tie. Net-negative on these themes, BUT all 12
regressions trace to a single root cause — the 5 fold-variants
helpers (_hyphen, _numeral, _accent, _honorific, _brit) live inside
_title_query_tokens at query.py:288-325 and providence_query
lazy-imports the WRAPPER without lifting the fold helpers.
Same gap manifests two ways:
- Wrong primary (5): Dr Who → pathology; Albert/Ahmed/Alaric the
third/first → wrong articles; Casa Batlló → error
- STRICT → HYBRID on correct primary (7): the verifier's Rule 8
title-overlap check calls the SAME _title_query_tokens —
without folds, "Andre-Marie" (claim) and "André-Marie" (title)
are distinct tokens, overlap fails, audit_mode demotes
Path A v3 surfaces: lift the fold-variants stack to _text_norm.py,
re-export from query.py, drop the lazy-imports in source_roles.py +
retrieval_routes.py. ~250 LOC moved + ~50 LOC import-rewrites,
half-day. Lower risk than v1 (pure code motion, helpers are
identical between paths).
Themes deliberately skipped this round (need their own gates ported
separately): quantifier_subset, metacog_subset, warrant_chain_probe,
es, fr. Re-bench AFTER v3 lands.
Also commits bench/legacy_vs_providence_bench.py + the result JSONL
so the regression set is reproducible.
Records the 2026-05-31 attempt at Path A (port the 5 reranks,
re-bench) and the result: smoke score went DOWN from 3/5 to 1/5
with reranks wired in. Root cause: legacy's rerank multipliers are
tuned against legacy's candidate-set shape (over_fetch=32, per-shard
parallel routes, body-density baked in earlier), not against my
multi_route fan-out's shape (per_route_limit=top_k*4, post-merge
candidates, filter-then-rerank instead of filter-during-route).
Helpers stayed in tree as importable building blocks (commit
d099995). Wire-up was reverted before commit so user surface is
unchanged.
Four v2 directions surfaced and documented for the future
investigation:
1. Match legacy's oversample factor (top_k*8+ or over_fetch=32)
2. Apply body-density filter BEFORE rerank cascade, at source
3. Rerun reranks on per-shard route output before final merge
4. Synonym expansion at retrieval time, not just filter time
None are blockers individually but each is a focused investigation.
The honest takeaway: legacy query()'s rerank pipeline is not a
"library of multipliers you compose in order" — it's a tightly
coupled cascade where each stage's tuning depends on what the
previous stages emitted. Collapsing it requires understanding
those couplings, not just lifting the helpers.
Documents the work shipped this session (Phase 1 foundation, 10
commits 9ba6317..20faae0) and the blocker that stops Phase 2: the
multi_route pipeline regresses on 2/5 smoke fixture questions
(Mercury Seven → Sam T. Beddingfield; dinosaurs extinct →
Paul Austin Kelly) because Phase 1 hasn't ported the 5 downstream
rerank stages legacy query() uses to suppress noisy phrase-route
hits — body-density check, body-coverage sqrt rerank, source-role
rerank, title-purity rerank, ordered-token-match rerank.
Path B chosen 2026-05-31: stop here, leave multi_route off by
default (which IS default — policy=None preserves body-only
behavior). Phase 1 foundation stays in tree as future-ready
infrastructure; user surface unchanged.
Path A (port the 5 stages, multi-day effort) reserved for a
future focused session. Until then, legacy query() keeps
producing the same answers it always has, INCLUDING the wrong
"Edwina" pick on dinosaur Q5. The slim-FTS5 cloud path already
fixes that bug for `arborist cloud query` / `arborist
corpus-query` callers (proven 5/5 source parity in d9fb6a9).
Bumps Next ID 000072 → 000073.
The three "92 claim_pack docs" tags were drifting against shard 000.db's
21 rows because the harness only counted one shard, but the doc prose
("#000031 closed at 92") meant the corpus total (21+16+38+17 across
genesis shards 000-003).
Two-line fix path: extend the harness to sum across all ???.db shards
via a `*:` prefix (e.g. `*:documents?source_type=claim_pack`), then
prefix the three drifted tags. Aligns the harness scope with the
semantic scope of the claim instead of forcing the claim to shrink to
one shard.
The `*:` glob:
- Matches `[0-9][0-9][0-9].db` basenames only (operator sidecars
qa.db / snapshots.db / selfmodel-chain.db skipped)
- Skips shards lacking the named table (schema-version tolerance)
- Returns _DB_MISSING when no genesis shard exists (CI / fresh-
checkout skip semantic preserved)
- Returns _TABLE_MISSING when no contributing shard has the table
Documented in ticket-000044 §3.4 + a third example showing the new
syntax. Diagnosis credit to a sub-agent investigation that confirmed
zero eviction/falsification audit events on claim_packs — the data is
intact; the harness was just single-shard.
Sibling to #000070 on the second axis of Joseph (@TrudoJo)'s spec.
Joseph posted the bridge-framework extension 6h after the 6-dim
spatial-ontology: chain A has its own world rules, chain B has its
own world rules, a bridge exists only where a deterministic
agreement layer can be proven. Three kinds: Agreement, Translation,
Embassy.
Phase 1 is doc-only -- the primary contribution is a retroactive-
validation appendix naming three already-shipped arborist primitives
as Joseph-bridge-shape patterns one-for-one:
#000066 graft-receipt mode = Agreement bridge
#000056 Operation Sandwich = Translation bridge (opus-mt hash
IS the translator_id)
#000059 graveyard (proposed) = Embassy bridge (POINTER-LINKED,
banner-labelled, never promotes)
The framework is not a new requirement on arborist. It is a unifying
spec that names what we already do in 3+ places, plus surfaces a
fourth pattern (bridge atlas for multi-instance federation with
potentially-different governance_policy_hash peers) as genuinely new
work for arborist/mesh/.
Bundle:
- docs/tickets/ticket-000071-world-bridge-grammar.md (~440 lines)
- docs/TICKETS.md index row + Next ID 000071 -> 000072
- docs/tickets/ticket-000070-spatial-anchor-pi-w-object.md §6 gains
one-line cross-reference to #000071 as adjacent axis
- Joseph memory extended to cover both frameworks
Phase 2/3/4 deferred to separate tickets. Hard constraints: A1-A3,
no audit_mode token, no SQL change at Phase 1, sovereignty rule.
Status: open, awaiting dav1d Phase 1 review + fox go/no-go.
Updates the spatial-anchor ticket so the pre-review empirical bench
(landed in 55b651f) reads as a self-contained dav1d review packet:
- Status line: notes bench shipped in 55b651f
- §7a Pre-review empirical evidence: expanded with reproducibility
notes (RNG seed 0xa8c90e551fd34427, wall ~2s, stdlib-only), the
five bench purposes mapped to open questions, full headline-results
table, and per-headline "what each one buys dav1d" prose for Q1
through Q9.
- §3 Land sequence: step 1 marked done in 55b651f with explicit
pre-review-bench scope; step 2 annotates that 5 of 10 open
questions now resolve by measurement.
- docs/TICKETS.md index row: leading status cell now opens with the
bench headline + per-section numbers so the index reads as the
dav1d handoff summary, not just "open · awaiting review".
No new code, no new commits to the bundle itself -- 55b651f remains
the bench's shipping commit. Pure documentation refresh.
New ticket for Joseph (@TrudoJo)'s 6-dim procedural spatial-anchor
framework as the first verifier kernel under the #000013 v7-W reserved
namespace. A single committed 32-byte SHA-256 hash deterministically
expands into six 32-byte regions H1..H6 via the HMAC-SHA-512 KDF already
shipped in arborist/substrate/anchor_prg.py (#000035); each region drives
one quantized object dimension under a fixed canonical mapper. Domain
separation from #000035 by dedicated spatial_anchor_seed published in
the v7-W manifest -- preserves #000035's KAT freeze + dav1d 2026-05-11
final review intact.
Bundle:
- docs/tickets/ticket-000070-spatial-anchor-pi-w-object.md (718 lines):
full spec with 8 design-choice subsections, working Python sketch,
12-NOT scope boundaries, 8-row cross-references, five-step deletions,
10 open questions for dav1d.
- bench/spatial_anchor_validation.py: pure-stdlib pre-review evidence
(~2s, RNG-seed-pinned, reproducible). Five benches: avalanche,
cell-distribution uniformity, collision vs birthday-bound, cross-region
independence, domain separation.
- bench/spatial_anchor_validation_results.md: report from first run.
- Makefile: 'make bench-spatial-anchor [SPATIAL_N=N]' target + PHONY.
- docs/TICKETS.md: index row + Next ID 000070 -> 000071.
Bench headlines (N=10000):
- Avalanche mean 767.85 bits (PRF null 768, z=-0.49) -> PASS
- Cell-distribution chi^2 |z|<1 at L=2,3,4 -> PASS
- Birthday-bound ratio obs/exp 0.989/1.038 at populated L -> PASS
- Cross-region Pearson all 15 pairs < 2sigma -> PASS
- Domain separation Arm A 767.91 / Arm B exact 0 collision -> PASS
Five of dav1d's ten open questions (Q1 seed source, Q2 segmentation,
Q3 position mapper, Q8 endianness, Q9 KAT adversarial vectors) now
resolve with measurements rather than appeals to PRF authority.
Q4/Q5/Q6/Q7/Q10 remain non-empirical design decisions.
Status: open, awaiting dav1d review + fox go/no-go. No registry slot
booked, no substrate-paper amendment landed, no kernel module created.
- make crawl-ingest writes to one central crawl db (CRAWL_DB, default
~/.arborist/crawl/web.db) instead of per-domain shards in the
peer-shared main dir: keeps locally-crawled content out of peer
sharing by default and a growing domain set under SQLite's 10-attach
cap (Makefile, docs/crawler.md).
- arborist query auto-includes the local crawl db (query() gains
extra_shards; CLI --include-shard / --no-crawl-db, default-on when
web.db exists). Fix latent --db single-file query AttributeError
(cli.py). Persist used / used_pointer_ids + retrieval_purity into
merkle_proof so read-only consumers can see which chunks fed the
answer (qa/query.py).
- arborist.read: read-only seam for dashboards / verifiers; on a
multi-source context root surface the real primary source instead of
the opaque corpus://multi-source sentinel (read.py). Backs the
arborist-viz Merkle Command Center (#000069).
- tests for extra_shards, the CLI crawl-db resolver, and the read seam.
Phase 2 — bench instrumentation + measurement run
bench/qa_sweep.py picks up the answerability sidecar projection per row
(answerability_fired, answerability_confidence, answerability_denial_
pattern, answerability_answer_type, answerability_candidate_count) and
aggregates per-mode (answerability_fires + S/M/W confidence breakdown)
into a new column in the markdown summary table.
Measurement run on bench/qa_results/phase2-sidecar-on/2026-05-27T14-
16-22Z (76 questions × n=3 × claim_lattice × Hermes-3-8B × tail layout,
228 runs). Headline:
sidecar fires 2/228 (0.88%)
confidence dist 2 strong / 0 medium / 0 weak
precision 100% (2/2 fires were the Ballestrini fixture)
recall on Ballestrini 2/3 across n=3 (third run model extracted
correctly -> sidecar silent,
correct behavior)
false positives 0/226 non-Ballestrini runs
verifier verdict both fires labeled STRICT by the binary
verifier (the verifier-blind class, exactly
as predicted)
Detection rule's three-clause conjunction (denial + extraction-shape +
candidate proximity near cleaned subject tokens) is operating at the
precision floor. The strong-confidence-only firing pattern is what
calibrates Phase 3's demote threshold.
Phase 3 — opt-in demote flag (default OFF per Dav1d Phase 4 NO-GO)
arborist/qa/keys.py: answerability_demote_enabled added to
_VERIFIER_POLICY_FIELDS so flipping the flag partitions cache via
verifier_policy_hash. Justification: when on, the rendered audit_mode
changes (EVIDENCE-WARRANTED -> EVIDENCE-MISSED-PARTIAL), which IS a
verifier-output property; verifier hash must move accordingly. The
other answerability_* fields stay governance-only (sidecar
diagnostic, no audit_mode mutation).
arborist/cli.py:_render_audit_label extended with answerability +
demote_enabled kwargs. Logic:
demote_triggers = (
demote_enabled
and answerability["answerability_warning"] is True
and answerability["confidence_class"] in ("strong", "medium")
)
lattice modes:
EVIDENCE-WARRANTED -> EVIDENCE-MISSED-PARTIAL (rung transition)
POINTER-LINKED / ANCHOR-WARRANTED -> "rung · missed-answer"
(tail tag; rung itself already
signals degradation)
non-lattice modes (quote/span/entity/paraphrase):
audit_mode token unchanged + "· missed-answer" tail tag
weak confidence: NEVER demotes (Phase 2 saw zero weak fires on real
failures; reserved for future expanded detection ladder)
CLI flag --demote-on-missed-answer on both `arborist query` and
`arborist ask`, default OFF. Flows into call_policy[
"answerability_demote_enabled"] and through to result[
"answerability_demote_enabled"] so the renderer reads it without
needing the policy dict.
End-to-end verified live: 4 fresh Hermes-3-8B runs with --demote-on-
missed-answer on `songs by veronica ballestrini`, all 4 rendered
EVIDENCE-MISSED-PARTIAL · via claim_lattice (Hermes hit the failure
mode in all 4, sidecar fired strong, demote logic transformed the
label).
Phase 4 (default flip to demote-on) — NO-GO per Dav1d 2026-05-27 §3.4:
"a false sidecar warning is tolerable; a false audit-label demotion
can damage trust in correct abstentions." Phase 2 precision is 100%
but n=2 fires is too few samples to claim precision floor empirically.
Default flip blocks on wider bench + human spot-check of the warnings.
Tests: 47 total (36 Phase 1 + 11 new Phase 3 covering hash partitioning
discipline + render-label projection across all four rung/confidence
matrices). Full suite 2794 passed (delta +22 from prior 2772).
Bench output (bench/qa_results/phase2-sidecar-on/) intentionally not
committed — bench/qa_results/ is gitignored per existing convention;
the ticket carries the headline numbers + path for re-inspection.
Adds a deterministic read-only sidecar to detect a class of failure the
binary verifier is structurally blind to:
Evidence contains the answer.
Model says the evidence does not contain the answer.
Verifier sees no unsupported positive claim -> marks run clean.
User receives a false negative under EVIDENCE-WARRANTED.
The motivating case: "songs by veronica ballestrini" against the 2010
Wikipedia corpus. Hermes-3-8B under user_payload_layout=tail returned
"the specific songs by her are not mentioned in the provided evidence
blocks" when evidence E2 literally contained "Amazing", "Out There
Somewhere", "Fascinated", "What's Up With That", "Don't Say". Verifier
correctly returned EVIDENCE-WARRANTED 2/2 because the existing layered
verifier (quote / span / entity / paraphrase + Rule 8 title-relevance +
Rule 9 subject-tokens-absent + claim-count ceiling) guards unsupported
*presence*, has no hook for unsupported *absence*.
Layout fixes attention placement on the specific instance (the 5/27
n=3x75q bench confirms bookend/per_chunk recover Ballestrini); layout
alone cannot close the class -- adversarial phrasing or a bigger prompt
resurfaces the failure under any layout. The right substrate move is to
falsify "not mentioned" as a testable claim.
Detection rule (three-clause conjunction, all must fire):
A. Denial pattern in answer (sealed v1 phrase list: "not mentioned",
"not provided", "the evidence does not say", "does not mention",
"no specific", "no evidence", "cannot determine from the provided
evidence", "is not stated", "is not specified"). Casefolded +
whitespace-normalized substring match.
B. Question is extraction/list-shaped. Either a surface cue ("songs
by", "works by", "books by", "who wrote", "who composed", "what
year", "list of", "name all", ...) matches, OR the existing
arborist.qa.quantifier classifier returns intensity in {ALL,
COMPREHENSIVE, OPEN_REQUEST, MANY, PLURAL}.
C. Evidence contains candidate spans matching the answer_type within
a proximity window (default 600 chars) of cleaned subject tokens.
Candidate kinds aligned to answer_type:
title_like -> quoted_string, title_case_span, comma_list_item
person -> title_case_span
date -> year, date
Hardenings folded in from the 2026-05-27 Dav1d de-novo review:
1. Subject tokens strip cue/relation/stop words. For "songs by
veronica ballestrini" the cleaned subject is ["veronica",
"ballestrini"], NOT all four tokens. Without this the guard
false-triggers on "Harvard University" or "New York" near
proper-noun subjects.
2. Answer-type alignment. Candidate span kind must match query type
so "songs by John Smith" + evidence about Harvard/NY does not
strong-trigger.
3. Confidence class is deterministic (weak | medium | strong), not
boolean. Strong requires quoted_string near exact subject mention
+ multiple type-matched candidates. Phase 3 demote will gate on
confidence_class.
4. Cap output at 10 candidates (the per_chunk-quote-inflation
lesson). Prevents the guard becoming another claim amplifier.
5. Offsets are offset_start + offset_end + offset_basis=
"evidence_object_text", never an ambiguous single offset.
6. Cache-hit path returns answerability: None. Cached records do not
carry the evidence_map, only the rendered sources summary, so the
sidecar cannot recompute candidate spans without re-running
retrieval. Operators wanting fresh diagnostics use --burn.
7. Phase 1 stays out of verifier_policy_hash. The
answerability_sidecar_enabled / answerability_threshold /
denial_patterns_version / extraction_cues_version fields fold
into governance_policy_hash only. Phase 3 demote flag
(answerability_demote_enabled, default False) will move the
verifier hash WHEN ON because it changes the rendered audit_mode
(EVIDENCE-WARRANTED -> EVIDENCE-MISSED-PARTIAL).
Sidecar discipline (matches arborist.qa.inspect.diagnose_* sister
functions deflection / coherence / title_relevance):
- no model calls (no LLM-as-judge, no NLI, no translation)
- no audit_events writes
- no providence_cache writes
- no answer text mutation
- no claim promotion -- the trigger conjunction makes promotion
structurally impossible (only fires on denial answers)
- byte-deterministic: same (question, answer, evidence, policy) ->
same output every time
Result-dict integration: result["answerability"] is None when the
guard did not fire, or a structured diagnostic dict when it did
(diagnostic_version, confidence_class, triggered_clauses,
denial_pattern_matched, extraction_cue_matched, extraction_shape,
answer_type, subject_tokens, candidate_count, threshold,
missed_answer_candidate_spans). Read by bench_qa (Phase 2 will add
warning-count aggregate to bench rows) and CLI render.
Three return points carry the key:
- miss-path (full retrieval + verify): computed from evidence_map
- cache-hit: None (Dav1d cache-hit recompute discipline -- evidence
not stored, recompute requires re-retrieval)
- reject-broad early-return: None (no evidence examined)
Tests: 36 new pinning the three-clause logic, positive (Ballestrini)
regression, negative control (John Smith + Harvard/NY), each-clause-
alone silence, schema integrity, byte-determinism, sidecar-disabled
short-circuit, dict-shaped evidence support. All pass; existing
inspect tests (60) all pass.
End-to-end verified live via the CLI on the real corpus (2010 ~/.arborist
/shards): 3 fresh Hermes-3-8B runs on "songs by veronica ballestrini",
run 1 hit the failure (sidecar fired with confidence: strong, 351
candidates, denial: "not mentioned"), runs 2-3 model extracted
correctly and sidecar correctly silent.
Phase 2 (bench + threshold tuning) and Phase 3 (opt-in demote flag)
are open as follow-ups. Per Dav1d: NO-GO on default demote-on until
benchmark + human spot-check confirms low false-positive rate.
Full spec in docs/tickets/ticket-000068-verifier-blind-missed-answer-
guard.md (post-review hardenings section at top names the seven
load-bearing changes from the Dav1d 2026-05-27 review).
2026-05-26 19:47 UTC. ~94 min wall. 3,468,226 globally-unique docs +
6,235,588 chunks + 90,592,990 edges + 3,468,403 audit events re-routed
from non-deterministic spray-by-ingest-order layout to canonical
content-hash M=4 layout (shard_idx = int(document_root[:8], 16) % 4).
Final state:
per-shard doc uniformity within ±0.04% (theoretical max ±0.05% for
first-32-bit SHA-256 prefix)
audit chain consolidated to shard 000 via Option A (re-sorted by ts,
re-chained; bodies preserved unchanged; tail event type=reshard
carries plan+result body, hash 8da3aa19…)
on-disk sizes: 000=11.0 / 001=8.8 / 002=8.8 / 003=8.8 GB
validation: chunks delta 176 (0.003%) + edges delta 547 (0.0006%)
are cross-shard dupes from re-ingest history, collapsed by
INSERT OR IGNORE; within the 1% tolerance gate
smoke queries: Barack Obama / YouTube / Albert Einstein all
returned proper evidence from correct (hash-routed) shards
chain-check-shards: 0 breaks on every shard
Two mid-flight defect fixes (also committed):
04edff7: derivations.src_root FK guard fired on legitimately
cross-shard refs → writer connection PRAGMA foreign_keys = OFF
c86d5ac: WAL accumulated ~37 GB across passes (SQLite auto-checkpoint
blocked by open reader cursors) → _checkpoint_truncate between
phases. Production migration was rescued mid-flight by manual
sibling-connection wal_checkpoint(TRUNCATE) freeing 27.7 GB.
Closes#000065 in both the index and the per-ticket file. The
per-ticket design doc stays open as a design-log artifact (its
content is still the right reference for the next reshard / for
graft mode #000066).
Follow-on tracked as tasks #44–#47:
#44 re-pack post-reshard shards into DO Spaces (current bucket
packs are stale, still in pre-reshard topology)
#45 verify bucket pack hydration is deterministic against the new
content-hash layout
#46 genesis a fresh peer on 3090-ai.foxhop.net from cloud — first
real SPV-wallet end-to-end test
#47 retire stale pre-reshard bucket packs after #46 confirms
Scaffold-only ticket. Captures the architecture for taking the
#000061 cold-pack format and adding a second mode beside hydrate:
overlay an existing pack onto a populated shard set ("graft").
Surfaced while running the #000065 reshard cutover and fox extended
the design: each pack carries a `corpus_name` field
(wikipedia-2010, wikipedia-current, arxiv-cs, ...), making
`arborist cold graft wikipedia-current` feel like `apt install`.
Three concerns analysed:
doc/chunk/edge overlay trivial (INSERT OR IGNORE on content-
addressed PKs collapses dupes)
FTS5 overlay trivial (new chunk_ids → new fts rows)
audit chain overlay the only hard part — three approaches:
A graft receipt (chosen): one event in
host chain carrying pack_hash +
event_count + first/last hashes; pack
file is the durable witness; zero
schema cost; aligned with v8/v9
witness pattern
B re-chain everything: rejected — graft
is frequent so invalidating external
refs is wrong tradeoff (different
story from the one-time reshard)
C chain forest with chain_id col: right
answer when graft dominates lifecycle
but premature now
Long-game payoff: mesh-peer-corpus-merge. Two peers diverge over a
partition, each carries packs the other lacks, reconciliation =
exchange + graft what's missing. Makes "mesh of arborists"
coherent rather than "fleet of arborists."
Scaffold gated on (a) #000065 lands+stabilises, (b) a second
corpus exists to graft, (c) at least two peers want to exchange.
No code until then; the design lock is what the ticket buys.
Index bumped Next ID 000066→000067.
Pinned the canonical shard count at M = 4 based on real-Wikipedia
ingest + query benchmark (bench/shard_count_sweep.py). Captured the
"when does SQLite stop being the right substrate" decision tree so
future operators know what bench would justify a fork or replacement.
Bench numbers (Wikipedia 2003 cur dump, 2000 docs, 4 cells of
M ∈ {1, 2, 4, 8}, 50 FTS queries per cell):
M chunks/s q_p50_ms q_p99_ms attach_ms
1 3,648 0.05 0.20 1.61
2 5,649 0.03 0.19 3.24
4 6,405 0.07 0.30 9.00
8 6,959 0.03 0.28 9.65
Key observations:
- M=1→M=2 is the biggest ingest win (+55%). Most gain happens there.
- M=2→M=4 is +13%. M=4→M=8 is only +9% — diminishing returns.
- Real wikitext canonicalization is per-worker Python CPU bound, not
SQLite-writer-lock bound. More shards don't unlock more CPU.
- Query p50/p99 is flat across M within noise (50 queries small).
- ATTACH cost grows linearly: 1.6 / 3.2 / 9.0 / 9.7 ms.
Why M=4 specifically:
- Captures 92% of peak ingest throughput (6,405 / 6,959).
- 6 ATTACH slots free under SQLite's 10 ceiling for aux DBs
(qa.db, snapshots.db, selfmodel-chain.db, crawl_*.db, future
mesh_*.db) — comfortable headroom. M=8 leaves only 2 slots.
- Mobile-tolerable: phone NAND attach is 5-10x slower than NVMe;
M=4 = 45-90 ms cold start (instant), M=8 = 50-100 ms (sluggish
with no headroom).
- Matches fox's current 4-shard layout = cheapest migration.
Decision tree for when SQLite stops being right (full text in
ticket §"When the SQLite-default substrate stops being right"):
A. ATTACH ceiling pressure (auxiliary DBs grow past 5) → bench
forked SQLite with SQLITE_MAX_ATTACHED=125, M ∈ {16, 32, 64};
if attach cost stays linear past M=10, fork viable but pays
permanent "no longer stock sqlite3" tax.
B. Ingest hits >10k chunks/s sustained ceiling → first tune
page_size / WAL checkpoint / mmap_size / synchronous. If
tuning gets 2-5x, stay on SQLite. If still ceiling-limited,
candidates: DuckDB (columnar, MVCC, FTS), libmdbx (B+tree no
FTS; we'd build it). In-house DB rejected without specific
failure of those.
C. Federation needs multi-writer-same-shard → SQLite writer-lock
serializes peers, becomes federation bottleneck. First try
leader-election (single-writer-per-shard with WAL replication
to followers). If true multi-writer required, SQLite is wrong;
candidates: FoundationDB, CRDT-on-KV-store. DuckDB does NOT
solve this — its MVCC is single-process.
Honest verdict: for current arborist workload (single-writer-per-
shard, read-mostly federation), stock python3 sqlite3 is the right
substrate. None of A/B/C are close to firing. The bench discipline
exists to know what to measure when something changes.
bench/results/shard-count-sweep-2026-05-26T16-20-48Z.csv (synthetic
baseline) + 2026-05-26T16-31-34Z.csv (real Wikipedia) committed as
the load-bearing measurement for the M=4 choice.
fox caught the imprecision: "the shards are teleported if we fix this
because we know the shard a doc will end up in."
Re-ingest = re-read source files, re-canonicalize, re-chunk, re-hash.
Hours-to-days from Wikipedia dumps.
What's actually needed = content-addressed rebalance. Every row in
the corpus is already addressed by content (document_root, leaf_hash,
merkle_nodes.hash, audit_event_hash). Migration reads existing rows,
computes new_shard_idx via the routing function, INSERTs into the new
M shards. No source-file work, no canonicalization, no LLM.
Updated the ticket's "Migration story" section with the per-table
teleport recipe:
documents: move by document_root → new shard
chunks (+content): follow doc to its new shard, re-assign chunk_id
chunks_fts: rebuild per new shard from chunks.content
merkle_nodes: move with document_root
edges: move by src_root
concept_relations, move by content-derived parent
derivations,
providence_cache,
citation_aliases,
term_aliases:
audit_events: ALL → shard 000 (canonical), re-numbered + re-hashed
snapshots, mesh_*: canonical shard 000
Audit chain handling: chose Option A (canonical shard 000) over
Option B (split-by-subject_root). Preserves global event ordering,
re-hashing happens once at migration time. The alternative loses
cross-shard event ordering for falsification + replay reasoning.
Estimated wall: ~20-40 min I/O-bound on current corpus
(35 GB across 4 shards → ~35 GB across 8 shards, moves once).
Hard discipline added: move into ~/.arborist/shards.v2/, keep
originals as rollback, atomic dir-swap when verified. Mistake in
routing function = corrupted federation, recoverable only by going
back to the originals.
Surfaced while sizing #000061's federation story. Today shard count
conflates two roles:
- producer ingest parallelism (wants = vCPU count)
- consumer ATTACH fan-out (capped at SQLITE_MAX_ATTACHED=10)
On default Python sqlite3, the ATTACH ceiling is 10 and can't be
raised without a custom sqlite3 build (which violates CLAUDE.md's
"python3 + venv + sqlite3 is enough" property). Producer with 16
shards → consumer fails to attach the 11th, federation silently
breaks.
Design: introduce M = canonical shard count (corpus-wide constant,
default 8) decoupled from N = ingest workers. Document → shard
assignment becomes content-deterministic:
shard_idx = int(document_root[:8], 16) % M
Same input → same output across every peer. Today's "spray by ingest
order" is non-deterministic across peers — two peers re-ingesting
the same corpus put the same document_root in different shards. That
weakens federation more than it should.
Migration hard-constraint (fox: "this implies we will need to
reprocess all our data into shards"): re-ingest required. Current
layout is sprayed by ingest order; post-ticket is sprayed by content
hash. Two layouts are incompatible by construction. Captured in
ticket §Migration as the load-bearing operational note.
Phases laid out (0-4: design lock → read path → ingest path →
pack-restore → corpus migration tool). Open audit-chain re-numbering
question (per-shard event_hash chains break when rows rebalance
across shards). Kept as one ticket — do-not-proliferate.
Scaffold only. No code yet — design lock first.
Next ID bumped to 000066.
Dav1d's reviews of #000061 (Response A + Response B/FINAL in
~/Downloads, 2026-05-26) flagged a long list of items — most already
shipped in the SPV-split work. Three were genuine gaps worth folding
into #000061 before close:
Gap 1: manifest/latest pointer for new-peer discovery.
A fresh peer doing `cold list` got a list of metadata-pack hashes
but no obvious "which one is current for shard X." Added
get_latest_pointer + update_latest_pointer to the backend ABC.
push_pack writes manifest/latest.json on every successful metadata
pack push (read-modify-write keyed by snapshot_root). Mutable
pointer; content addressing of the packs themselves preserves the
trust root. Last-writer-wins on contention.
Gap 2: license_class field + producer-side refuse for public buckets.
Maps documents.source_type to a license bucket (wikipedia_cur /
textbook_tex → public_redistributable; html / grok / vcs → unknown;
anything else → unknown). Strictness order: public < unknown <
private. compute_shard_license_class() walks DISTINCT source_type
in documents. push_pack now refuses to upload if the shard's
strictest license is more restrictive than the operator's
allow_license_class (default: public_redistributable). The
metadata pack's manifest carries _license_class so consumers /
auditors can see the producer's classification without inspecting
source documents. ValueError on refusal — the bucket ACL is the
operator's call, but arborist refuses to participate in a
licensing/membership leak unless explicitly opted in.
Gap 3: cold_pending table for resumable uploads.
Killed mid-upload, push_pack left orphan multi-GB tempfiles in
/tmp with no DB trace. Added schema:
CREATE TABLE cold_pending (
tempfile_path TEXT PRIMARY KEY,
pack_hash TEXT NOT NULL,
kind TEXT NOT NULL,
backend_endpoint TEXT NOT NULL,
backend_bucket TEXT NOT NULL,
object_key TEXT NOT NULL,
started_at INTEGER NOT NULL,
state TEXT NOT NULL DEFAULT 'pending'
);
push_pack INSERTs a row before each upload + DELETEs on success.
A killed process leaves the row pointing at the orphan tempfile;
a recovery script (future) reads cold_pending, checks bucket for
the object, either deletes the row + tempfile (success was just
unreported) or re-uploads from the tempfile if it still exists.
Matches the same pattern as the audit chain — explicit state
rows beat inferring from chunks.content IS NULL.
Sibling tickets opened for the larger items the reviews flagged
(scaffold-only, no code; opening them captures the design in the
log without proliferating, per CLAUDE.md):
- #000063 Cold-object private-ciphertext mode (mesh-keyed object
keys for non-public corpora on public-read buckets). Needs mesh
group-key ABI + real non-public corpus before code.
- #000064 Cold-object operations toolkit (verify / diff / doctor /
repair-fts / gc-plan CLI + expanded audit-event taxonomy).
Bundled so the audit-event vocabulary gets one design pass.
5 new tests:
test_gap2_license_gate_refuses_unknown_class_to_public_bucket
test_gap2_license_class_in_metadata_manifest
test_gap1_latest_pointer_resolves_metadata_pack_per_snapshot
test_gap3_cold_pending_clears_on_successful_upload
test_gap3_cold_pending_records_inflight_upload
26 cold-object + 7 evict tests pass (33/33 green incl. boto3 wire).
Next ID bumped to 000065.
Live v3 SPV corpus run (bmq47x6t3) completed cleanly during this work.
Will report sizing + memory profile in the next message.
Three doc-only housekeeping items from Dav1d's de-novo reconciliation
of the architecture stack (no code changes; the active build is
#000061 cold-pack work, which is unaffected by this review):
1. Accepted-error formula. Dav1d corrected p_raw × (1-d)(1-r) to
p_raw × (1 - dr) where r is conditional on detection. Searched
docs/ and arborist/ — the wrong form does NOT appear in this
tree (it lives in the external recursive-truth-maintenance / RCO
manuscripts Dav1d cited). Nothing to fix here; recorded for the
manuscript authors.
2. #000060 H-ABCDEFG split. Folded the M/C/X axis split into the
harness scope: M (mechanism — does the substrate's plumbing work),
C (capability — does it improve task performance), X (external
adversarial — does it generalize outside author-designed fixtures).
ACCEPT requires clearing all three. Without X, the harness risks
self-validating benchmark theology — passing tests its own designers
picked. Tagged at fixture-definition time; aggregator emits per-axis
pass rates + combined ACCEPT verdict. Doesn't change the existing
BatteryResult row schema.
3. #000062 Mechanistic Witness — new scaffold-only ticket. Specifies
a content-addressed MechanisticWitnessRoot over (model_config,
capture_policy, contrastive_prompts, features/neurons, intervention
result, behavioral delta, safety policy) as a DIAGNOSTIC sidecar
feeding SelfModel + benchmark fixtures. Four hard guardrails:
diagnostic-only by default; sandbox intervention only; no production
steering without governance ACCEPT via #000060 M+C+X; feature labels
never become semantic proof. No code until a real falsifier use case
exists + guardrails are CLAUDE.md rules + #000060 harness gates
promotion. The dual-use risk (Pan et al. 2025: 0.1% MLP ablation
breaks refusal in 72B models) makes governance-first framing
load-bearing.
Next ID bumped to 000063.
No code change to arborist/. The in-flight v3 SPV corpus pack
(bmq47x6t3) continues unaffected.
v1 packs (chunks-only) were under-engineered: a new peer landing on
v1 packs would have chunk bodies indexed by leaf_hash but no documents
table, no audit chain, no merkle interior, no edges — couldn't actually
hydrate. fox: "isn't what I wanted you under engineered..."
v2 packs ship every load-bearing shard table alongside chunk bodies in
the same tar.zst:
manifest.jsonl # chunk catalog (unchanged)
tables/documents.jsonl # array-per-line columnar JSONL
tables/chunks.jsonl # without content column
tables/merkle_nodes.jsonl
tables/edges.jsonl # FAN-IN restructured
tables/audit_events.jsonl
tables/derivations.jsonl
tables/concept_relations.jsonl
tables/concept_token_idf.jsonl
tables/providence_cache.jsonl
tables/citation_aliases.jsonl
tables/term_aliases.jsonl
tables/snapshots.jsonl
tables/document_http_meta.jsonl
blobs/<hash[:2]>/<hash[2:]> # raw UTF-8 chunk bodies
Two compression strategies inside the pack:
1. Array-per-line JSONL ({"_columns": [...]} header line + ["v1","v2",...]
data lines) drops ~30% of uncompressed bytes vs object-per-row JSONL.
zstd recovers most of that on its own, but smaller uncompressed
footprint also speeds up stream-restore.
2. Edges fan-in restructure at pack-build time: 22M rows of
(src_root, edge_type, dst_root, dst_uri, anchor) → ~500k unique
(dst_uri, edge_type, anchor, dst_root) groups with src_roots as an
array. ~5-10x compressed savings on the dominant table. Reverses on
unpack into the per-edge live schema. Live queries unchanged.
NOT shipped (per-peer state): mesh_*, selfmodel_*, capital_ledger,
memory_*, controller_events, fork_score_branches, adapter_loss_reports,
falsifications, schema_meta, meta. NOT shipped (rebuildable): chunks_fts*,
documents_fts* — restored from chunks.content + documents.title on
unpack.
push_pack no longer appends `cold_pack_pushed` to the audit chain.
That event leaked into the next push's audit_events.jsonl dump and
broke the "two writers at the same corpus state produce identical
pack_hash" determinism property. The bucket/disc file IS the receipt;
the snapshot_root pinned inside the pack metadata binds it to a corpus
state. No load-bearing consumer of the audit row.
pull_pack restored to handle both v1 (chunks-only) and v2 (tables +
chunks) packs. For v2 it extracts tables/*.jsonl to a temp dir,
calls restore_shard_metadata (which INSERT OR IGNOREs into the live
schema and expands edges back to per-edge rows), then fills chunk
content for every leaf_hash in blobs/. Idempotent against populated
DBs (INSERT OR IGNORE all the way down). Self-cleaning temp dir.
Sizing measured 2026-05-26: ~2.1 GB per shard pack compressed (chunk
content 1.78 GB + metadata ~0.3 GB), ~8.5 GB total across 4 shards.
~20% more than v1 chunks-only for self-sufficient hydration.
24 cold-object + evict tests pass (+1 new test_push_pack_v2_hydrates_fresh_empty_db
that builds a pack from a populated DB and unpacks into a completely
empty DB to verify all tables restored). Full suite: 2558 passed,
28 skipped, 1 xfailed.
Ship arborist corpus state to new peers and DVD-R archival via
point-in-time tar.zst packs. One artifact serves both channels —
bucket+CDN delivery and physical-media archival.
Bucket holds packs only. Pack key = hash_leaf(manifest_bytes), so same
chunk set on two writers produces the same pack_hash and upload is
idempotent. Each pack pins the corpus snapshot_root it covers in audit
+ result body — packs are delayed snapshots, not live mirrors;
falsifications between repacks produce new pack_hashes.
stream_packs runs streaming zstd over tarfile, peeking compressed-buffer
size after each chunk via FLUSH_BLOCK (preserves dictionary). Default
cap 4_400_000_000 — 4.4 GB DVD-R safe-fit, ~6.5% buffer below the
4.7 GB marketing capacity to absorb ISO9660 overhead, growisofs
lead-in/lead-out, media variance, and drive-edge refusal. Each disc
fills to ~4.4 GB recorded data, not the ~1.5 GB an uncompressed cap
produced.
One backend class (S3CompatibleBackend via boto3 + endpoint_url) covers
AWS S3, DO Spaces, R2, B2, GCS S3-interop, MinIO. Optional dep
[object-store] = boto3>=1.34; dev extras pull moto for the wire test.
Voyeur: credentials via AWS_ACCESS_KEY_ID/_SECRET_ACCESS_KEY env or
~/.aws/credentials, never printed; only endpoint URL + bucket name
surface in logs.
CLI: arborist cold {pack,unpack,stats}. Makefile: cold-pack,
cold-pack-dvd (local-dir output for growisofs), cold-unpack, cold-stats.
Sizing for current shards (14.1M chunks, ~17 GB compressed): ~4 packs
at the default cap, ~\$0.34/mo DO Spaces storage, ~\$0.0001/fresh-peer
hydrate.
Always-on raw-UTF-8 leaf store (per ticket "Hard invariants") deferred
— packs-only for now, backfill later.
2557 passed, 28 skipped, 1 xfailed.
v1 of the same-model substrate-delta harness's non-jagged metric.
For one corpus title, surface-perturb its question (numeral / accent /
hyphen / honorific / amp / brit) preserving the referent, then ask
whether retrieval surfaces the SAME target for canonical vs perturbed
phrasing. J_norm = XOR disagreement rate @k (lower = less jagged);
graded mean |Δrank| catches rank instability the binary metric misses.
Pure query --dry-run: no LLM, no verifier, no judge, no n=3 noise, no
5pp floor — the recall_at_k discipline. Reuses recall_at_k.probe +
mine_questions._surface_variant. Feeds #000012 ForkScore
ΔJaggednessReduction. A-vs-C answer-quality arm already exists under
#000057 (control_ab/control_sweep) — not rebuilt. Curvature + LLM-arm
jaggedness delta remain open (ticket §8).
make bench-jaggedness JAGGED_LIMIT=40 JAGGED_K=8
fox: a graveyard that keeps every wrong answer forever grows unbounded
like cancer and is counter to constraint-optimization. Bake in the
bound:
- earn-to-enter (recurrence-gated) — a failure earns a tombstone only
when its canonical claim-family re-fails; one-offs dropped.
- fingerprints not transcripts (UTXO-set analogy) — store
canonical_claim_family|falsifier_root|failure_class, prune bulk text.
- decay/compact to steady-state — evicts like a surface, never a core.
- off the hot path — dedicated graveyard shard, bloom-filter burden
check only.
Gossip-group falsifier admission inherits the same discipline:
difficulty-adjusted stable-rate ingestion (BTC block-rate target) +
per-window budget (#000036), enforced in #000012/mesh. BTC lesson is
bounded self-regulating ingestion, not store-everything. If it cannot
be bounded to a steady state, the graveyard is not built.
Three tickets from the 2026-05-20 Dav1dPrometheus "Protocol-Layer AGI"
working report (held outside the repo; referenced not committed):
- #000058 cache_key_9 verifier-policy mandatory-vs-legible decision +
doc reconcile. Records the five-step-#1 correction: verifier fields
already fold into governance_policy_hash, so the 9th dim is audit
legibility not a correctness gap.
- #000059 admission discipline: claim-graveyard burden-shift +
self-providence quarantine (guards the existing ingest-self-providence
self-confirmation loop).
- #000060 H-ABCDEFG same-model substrate-delta harness (jaggedness +
curvature); curvature-aware ForkScore folded into #000012 Phase 2,
not spawned as a sibling.
Reconciles CLAUDE.md cache_key invariant (8-dim -> 8 + optional 9th).
Next ID 000058 -> 000061.
Opt-in third NLI inference backend (ARBORIST_NLI_BACKEND=tinygrad) in
qa/nli/shadow.py, parallel to torch/onnx-int8, behind the #000049
cage: shadow-only, never an audit_mode input, never auto-preempts the
proven path (guarded so it cannot regress torch/onnx). Loads the ONNX
export through tinygrad's frontend wrapped to the existing
model(**enc).logits contract so _nli_batch is byte-unchanged.
bench/scripts/nli_backend_ab.py + make bench-nli-backends: deterministic
A/B, gate is numerical agreement with the torch reference first,
latency second (a divergent engine = a different shadow signal = a
different nli_policy_hash). Instrument is honest — reports a requested
backend as unavailable rather than relabelling a fallback's numbers.
First CPU-smoke run already quantified that the deployed §7 #22 int8
export diverges Δmax≈0.42 from torch — the immunity property made
measurable, not a defect. Real tinygrad numbers pending a producer-box
run (tinygrad not an arborist dep; frontend op-coverage for the large
MNLI checkpoints unverified by design).
docs/onnx-vendor-capture-immunity.md: why the model-in-proof-path cage
makes the inference engine an interchangeable sidecar, never a trust
dependency — public-domain positioning capital. Indexed in CLAUDE.md.
Full suite 2498 passed (identical to baseline); 24/24 NLI tests green.
bench/control_ab.py + `make control-ab` (gated on judge-self-test
as a make dependency — instrument gate cannot be skipped). Same
model both arms; gold = target-article text by target_root; Arborist
[E…] scaffolding stripped (blinding — format can't betray the arm);
Arborist UNGROUNDED credited as honest abstention; hermetic Opus
judge; deterministic aggregate; self-auditing JSONL; threats-to-
validity printed in the report.
N=2 smoke: clean end-to-end, 0 JUDGE_ERROR — and already surfaced a
case AGAINST the treatment (solo correctly ABSTAINED; Arborist
HYBRID-WRONG). The instrument can falsify the Arborist value claim;
that is the point. n=2 proves nothing (report says so) — verdict
needs a real N.
fox ruled judge = Opus via `claude -p`. bench/judge.py:
hermetic (`env -u CLAUDECODE claude -p`, fresh process, context =
only (Q, answer, gold) — no arm label, no Arborist context, no
session), blinded-by-caller, reference-grounded against the fixed
gold (ignore parametric knowledge), structured via FINAL_VERDICT=
sentinel parsed LAST-match.
Instrument-before-experiment gate worked: first cut parsed
first-match over the model's chain-of-thought → 0/3 self-test. The
judge REASONED correctly; the parser was the defect (+ two bad test
fixtures, my error). Hardened (sentinel contract + fixed fixtures),
re-verified: `make judge-self-test` = 4/4 on known-verdict triples
via real claude -p. The make target is the precondition gate; no
control run trusts the judge until it passes.
Threat to validity recorded, not hidden: same model family judging;
mitigated (blind + no-stake + reference-grounded) not eliminated —
different-family SOTA cross-check is the only full removal.
Next: bench/control_ab.py + `make control-ab` (Hermes-solo vs
Arborist, gold=target-article text, blinded, judged) — NOT yet
built; no broken make target shipped for it.
fox: should the judge get the same docs, or the 8 retrieved/cited?
Neither. Per-arm retrieved docs (a) are undefined for Hermes-solo
(no retrieval → breaks A/B symmetry), (b) re-import circularity
(judging the treatment vs its own retrieval), (c) measure citation
self-consistency (Arborist's verifier already does that), not
truth. Judge's reference = the mined question's fixed ground-truth
target/gold, identical for both arms, independent of retrieval —
the ground-truth-carrying fixture supplies exactly this. Recorded a
distinct, non-headline faithfulness ablation (judge Arborist vs its
own context) so it is never conflated with the control delta.
Index row unchanged (skim surface; ticket body is the design log).
I over-applied a proof-path rule. "No LLM-as-judge" guards Arborist's
*verifier* (an LLM in the proof path = the circularity Arborist
kills; #000049's poor reliability is a proof-path-model constraint).
The control experiment's judge is external science — outside BOTH
arms, scoring post-hoc, touching neither system's internals — so it
cannot break the experiment. fox: use a strong SOTA model
(stronger than both systems under test), blinded + reference-
grounded + transcript-logged. This lifts the wrongly-imposed
atomic-fact-only limit (judge handles essay grounding; #000049's
exclusion was right for the proof path, wrong for external science);
deterministic gold-key demoted to a cheap atomic cross-check.
Still fox calls: fixture population, which SOTA judge (API
cost/egress), spend. Scaffold; no build past gate-1.
The review's strongest point (§4.1) + fox: every measurement so far
is WITHIN Arborist — no control arm. Folded into #000057 (not a new
ticket): control = Hermes-solo; judge external/deterministic/
paraphrase-tolerant via a mined gold-fact answer key (resolves
circularity + no-LLM-judge + paraphrase at once); headline metric =
fabrication-vs-honest-abstention delta, not raw accuracy (common-
knowledge fixture would falsely show null — parametric memory).
Smallest-proof reorder: v1 = single-shot Hermes-solo vs Arborist on
a fabrication-bait atomic-fact fixture; recursive-drift loop is v2,
gated on v1. Atomic-fact only; essay-grounding = #000049's unsolved
NLI, OUT. Still fox calls: fixture population, gold-key shapes,
spend. Scaffold; no build past gate-1.
The 2026-05-19 GPT-5.5 Hamming-framed review's ONE arborist-scoped,
ticket-worthy nugget: prove the Merkle-Providence-Reverse-RAG
whitepaper's headline claim (untracked evidence loss -> unbounded
recursive drift; witness-preserving state bounds it). Scaffold only,
awaiting fox go/no-go on scope.
Discipline encoded from the 2026-05-18 precedent (a grand target is
unfalsifiable until the instrument can resolve it — four hypotheses
died, only the deterministic mined-recall instrument broke it):
instrument-before-experiment, ONE task not eight, minimal ON/OFF
A/B, non-claims pinned (necessary substrate, NOT AGI).
Everything else in the review (rename, corpus hierarchy, IQ/talent,
ToE/Riemann/identity/geopolitics) deliberately NOT ticketed —
narrative/positioning, not arborist engineering; don't-proliferate.
Exactly one ticket. Next ID 000057 -> 000058 (same commit).
The fr-no-mask sweep (2026-05-18) corrected an over-pessimistic read:
French was reported 47% but that was the failed-mask artifact; honest
fr no-mask is 61%. §9 + index now carry the firmed n=1 table (EN 85%
· es-nomask 71% · es+mask 65% · fr-nomask 61% · fr+mask 47%), the
comparator-corrected verdict (sandwich is a ~0%→61-71% net win vs the
real no-support baseline; the −14/−24pp is vs native English, not a
regression), and the entity-mask lever's confirmed net-negative
status (default-OFF). Doc-only; no code change.
Three workstreams, full suite 2482 passed, experimental paths default-OFF.
#000055 — Windows quickstart without make
tasks.py (pure-stdlib runner) + make.bat shim + .gitattributes;
README Windows section rewritten. Quickstart needs only Python
3.10+ (no make/bzip2/curl/bash). Mirrors the Makefile quickstart
subset; drift-pinned by tests/test_tasks_runner.py.
#000001 §7 Phase 0 — deterministic cross-language guard
arborist/qa/crosslang.py: non-English signal (¿/¡/non-ASCII) + an
es function-word stoppack. Fail-closed to UNGROUNDED before
retrieval/LLM (mirrors the quantifier reject-DAG) when no content
token survives, else strips es stopwords from the retrieval query
only. English path byte-identical by construction. Default OFF
(crosslang_guard_enabled). Measured: the anarcocapitalismo field
case 10.4s -> 1.6s.
#000056 — Operation Sandwich (cross-language grounding)
arborist/qa/mt/: opus-mt es/fr/ru<->en, lazy per-pair memoised
singleton (fixes the 88%-engine-error concurrency defect),
manifest-pinned, [mt] extra; entity_mask wrapper. Sandwich =
translate query in (retrieval + LLM prompt) -> English answer ->
UNTOUCHED verifier grounds English-vs-English -> translate the
verified answer out as display-only (banner-labelled, zero
grounding). question_hash + verifier_policy_hash invariant; MT
engine identity binds into RetrievalPlan, not governance. CLI
--crosslang-translate / make XLANG_MT=1. Default OFF; entity_mask
default OFF (measured net-negative at bench scale). Fan-out bench
(bench/*.py): Spanish ~0% -> 71% grounded vs the real no-support
baseline; the round-trip predictor was tried and refuted; the
entity-mask lever failed at scale (corpus-title anchoring untried).
CLAUDE.md: cross-language bright-line convention + module map.
Pre-existing modified diagram files are intentionally excluded.
All 6 swept rerankers (bge-large/base, MiniLM-L-2/L-4/L-6/L-12)
false-fire on 4.5–21.8% of real bench-qa STRICT at their
candidate-bench fp=0 θ. The smallest θ that yields fp=0 on real
STRICT is BELOW the candidate-bench NEG max for every model —
meaning at the runtime-safe θ, catch on the 12 candidate-bench NEG
= 0/12 across the board.
Structural reason: real bench-qa STRICT answers have a much wider
score distribution (bge-large STRICT: min -2.20, p10 +2.99, p50
+5.80, p90 +7.54) than the tight contrived candidate-bench POS
band. The bottom 10% of legitimate STRICT score below where the
candidate-bench NEG cases sat. Distributions overlap heavily; no
threshold separates them.
This is the §3.2 mirror of #000049 §7 #27's recall-side walk-back —
clean candidate-bench → fails real-pipeline gate. Same diagnosis:
lexical-candidate selection + cross-encoder scoring + hard threshold
doesn't survive real-pipeline heterogeneity.
Verdict: relevance reranker CANNOT be promoted to a runtime
demotion-only veto on this design. demote_below_score stays null;
manifest gains runtime_viability block documenting the negative
result + the still-viable advisory soft-signal uses (render-tail,
multi-signal advisory, contrastive Q→A vs claim→source delta).
bench/scripts/relevance_shadow_sweep.py + bench/results/
relevance-shadow-sweep-pooled808.json committed. Production
verifier unchanged; advisory only; no audit_mode effect.
§3.1 diagnose_coherence (now 19 tests, +10 from the parallel session's 9):
- 3 more positive shapes (multi-sentence vacuous, named-entity circular,
grammar-term phrase_component_reuse).
- 5 xfail regression tests for SHAPES THAT FALSE-FIRE on real bench-qa
STRICT data (44/808 = 5.4% FP rate measured on the pooled n=1+3+5
STRICT answers). Each xfail names the exact shape + why it should
ideally be 'ok' + which rule needs tightening:
* 'The chemical symbol for gold is Au.' → vacuous (short predicate)
* 'Michael Jordan's Restaurant was a restaurant ... named after
Michael Jordan.' → circular (named-after re-use)
* 'The Western X was the western half of the X' → circular
* 'The name <Phrase> is a translation ... of the <derivative>' →
phrase_component_reuse (translation/etymology)
* claim-lattice [E1 | … …"] tails → vacuous (truncated bracket
fragment)
* 'The term <X>' → phrase_component_reuse (idiomatic English)
- 1 load-bearing real-traffic test: FP rate on 808-cell pooled STRICT
must stay ≤ 7% (current 5.4%) — fires loud if a future change
regresses it. Skips on fresh-checkout (bench/qa_results/ gitignored).
§3.2 ShadowRelevance (now 20 tests, +7 from the round-1 scaffold):
- Manifest tests for round-2 primary (bge-reranker-large), the size
spectrum coverage (50-560MB), the candidate-bench findings block
(biggest-within-family / not-across-families / deeper-not-better /
capacity-floor).
- Pair-kind distinction (question_answer vs claim_source recorded
separately for downstream telemetry / governance hashing).
- Batch-order preservation (_score_batch must return scores in input
order — load-bearing for downstream zip-back).
- Empty-input handling (Q empty, D empty, whitespace-only).
- Zionist-entity discriminator sanity (on-topic > off-topic logit).
- demote_below_score-stays-null invariant (the §7 #18→#27 discipline:
no hardcoded threshold; must come from a real-traffic shadow sweep).
Total: 101 passed + 6 xfailed (5 §3.1 regressions documented + 1 from
parallel session). The 5 xfails are the bench-maxing receipts — they
document EXACTLY which shapes §3.1 false-fires on, with the rule that
needs tightening named in each reason.
`make bench-qa BENCH_QA_N=3 BENCH_QA_LIMIT=5` (2026-05-13T14:24Z) on
the bench/qa_questions.txt set: 30/45 STRICT (67%), zero regressions
on mona lisa / capital of france / new london bridge (each 9/9 STRICT
across quote / pointer / lattice modes). The bench-qa-smoke n=1
flickers I saw ("dinosaurs" S→U, "soviet union" S→H) were LLM
stochasticity at n=1, not retrieval bugs. GNU-linux + python failures
look like genuine hard questions, not Phase-2 side effects.
End-to-end gap-close from the Phase 1 extractor. Five interlocking
fixes; live-verified that `what is a CPU?` → "Central processing unit"
at #1, `what is a GPU?` → "Graphics processing unit" at #1
EVIDENCE-WARRANTED 1/1; Mount Kilimanjaro / Soviet Union queries
unchanged (no regression).
(a) `synonym_expand` over-cap path is now rank-and-truncate by
source-frequency (descending) instead of hard-skip. CPU has 13
legitimate homonym expansions across the corpus; the prior
MAX_NEIGHBORS_PER_TOKEN=8 cap contributed *zero* expansion → no
canonical-article surfacing. Now: keep the 8 dominant by per-(token,
target) source-root count via _load_neighbor_source_freq.
(b) `_search_titles` orders by FTS5 bm25 ASC instead of LENGTH(title)
ASC on the FTS5-MATCH path. The length-asc tie-break was correct for
the 2026-05-02 "Back to the Future" LIKE-substring case but
counter-productive on FTS5 (tokenized; no substring junk; length-asc
preferred "Unit" / "Unite" / "B unit" over "Graphics processing
unit"). LIKE fallback keeps length-asc since the substring issue
persists there.
(c) `accept_tokens` (feeds title-search, core-keyword, title-rerank)
uses the synonym-expanded set instead of qtokens-only. The Phase 1
expansion existed but was only used in the FTS5 OR-fallback; satellite
articles saturated the budget before the canonical article entered.
(d) ARCHITECTURAL: `synonym_expand_strict()` (new — high-trust
evidence-kind subset: manual + manual_legacy + acronym_parens,
**excludes** link_reciprocity) for use in the multiplicative
`_rerank_by_title_purity` and as the source for `accept_tokens`.
Reciprocal-wikilink edges express *topical adjacency*, not synonymy
(a `Dinosaurs` page reciprocally links to `Curious George Brigade` →
edge that should not amplify retrieval); a multiplicative ranker over
them blows up. Strict view preserves the acronym-parens surfacing
(those edges ARE the phrase=expansion identity) while keeping
link-reciprocity to additive retrieval-route boosts via the broad
synonym_expand (still wired to `or_synonym_pool` for the FTS5 OR
fallback).
(e) Extractor regex tightened `[A-Z]{2,6}` → `[A-Z]{3,6}` and purged
~21K 2-letter acronym edges from shards. 2-letter acronyms (AI/ML/OS/
US/UK/IT/PC/TV) homonym-collide too often with common 2-letter QUERY
tokens like `go`/`is`/`am` — without this, "why did the dinosaurs go
extinct?" pulled Curious George Brigade via GO-acronym edges. The
high-value acronyms (CPU/GPU/RAM/DNA/FBI/WHO/…) all clear 3 chars.
Also: CLAUDE.md gains a "prefer existing ticket; only split for
Dav1d-review audience" discipline note (saved as feedback memory) —
this work is itself an example: would have been #000055 + #000056 +
#000057 under the prior pattern; instead extends #000054.
Suite: 2531 passed (no regression). bench-qa in flight separately.
All 4 manifest candidates clean-separate (12/12 NEG catch at 0/14 POS FP),
so 'reranker discriminates Zionist-entity-style mis-cites' is a property
of MS-MARCO-trained rerankers as a class — not the specific L-6 I picked
first. Ranked by separation margin (per §7 #18: separation beats raw):
ms-marco-electra-base +5.349 ← new primary
BAAI/bge-reranker-base +3.076
ms-marco-MiniLM-L-6-v2 +3.053 (previous primary, demoted)
ms-marco-MiniLM-L-12-v2 +2.663 (worst — deeper ≠ better)
Manifest primary moved to electra-base for the cushion. demote_below_score
STAYS null — clean candidate-bench thresholds don't predict real-pipeline
behavior (the §7 #18→#27 history is 6 verdict flips on the NLI side);
§3.2.2 step 2 (real-traffic shadow sweep on pooled bench-qa STRICT) is
what sets it. Expect a walk-back. Reranker still doesn't catch the
Kilimanjaro/Mount-Kenya recombination (aboutness ≠ truth-of-attribution;
that's #000049 territory). fox's bench-maxing correction applied.
arborist/qa/relevance/ — manifest pins cross-encoder/ms-marco-MiniLM-L-6-v2
(~80MB, Apache) as primary; alternates: L-12, BAAI/bge-reranker-base,
ms-marco-electra-base. demote_below_score=null on purpose — the
#000049 §7 #18→#27 discipline (proved 6× that clean-eval thresholds
don't transfer to bench-qa data) requires the threshold to be set by a
shadow sweep against pooled real STRICT, not by a literature number.
ShadowRelevance class mirrors ShadowNLI (lazy [nli]-extra import, cuda
auto-detect via ARBORIST_RELEVANCE_DEVICE or ARBORIST_NLI_DEVICE,
batched _score_batch, graceful degrade-to-available=False). Two surface
methods: check_question_answer (deflection / Q-A drift) and
check_claim_source (topic-collision mis-cite). 13 tests.
Sanity on the motivating field case (Zionist entity): ON-topic +9.96
vs OFF-topic -9.04 → 18-pt margin. Mona Lisa Q→A deflection: on +10.45
vs deflect +3.56 → ~7-pt margin. The model CLEANLY discriminates the
failure modes #000052 §1 named. It does NOT catch the
recombination-where-the-different-entity-clause-also-mentions-the-target
case (Kilimanjaro/Mount Kenya) — and that's the right architectural
split: aboutness (#000052 §3.2) and entailment (#000049 NLI) are
orthogonal axes; the Kilimanjaro recombination case needs the semantic
candidate selector (#000050/#000051 vec hybrid).
Remaining: build candidate-bench eval (~20-30 deflection + mis-cite
fixtures), shadow-sweep θ over pooled bench-qa STRICT (expect another
walk-back per the #000049 lesson), recall-side realism check, then
fox+dav1d sign-off. Still SHADOW; production verifier unchanged.
`arborist/concepts/extract.py:acronym_parens_synonym` — new
corpus-agnostic extractor. Scans each doc's lead chunk (first 4000
chars) for `<Multi-Word Phrase> (ACRO)` where the all-caps acronym's
letters strictly match the content-word initials of the phrase, in
order, after function-word filtering. Emits bidirectional synonym
edges between the lowercased acronym and each ≥3-char content token
of the phrase, evidence_kind="acronym_parens", anchored to that doc's
document_root. Idempotent like link_reciprocity_synonym.
Why this complements link_reciprocity: Wikipedia represents
abbreviation→expansion as a one-way *redirect* (CPU →
Central processing unit), which the ingest does not record as an
edge — so the existing reciprocal-link extractor never learned the
relation. The relation IS in body text by near-universal convention
("Central processing unit (CPU) is..."), which this extractor reads.
Corpus-agnostic: HTML, blogs, textbooks benefit equally.
Conservative: strict 1:1 acronym-to-atom match (rejects HTTP-shape,
where letters land mid-word), function words filtered, repeated
definitions deduped per doc, ≥3-char target floor. 8 new tests
covering CPU bidirectional emit, RAM idempotency, FBI function-word
filter, HTTP length-mismatch reject, XYZ initial-mismatch reject,
ROM hyphenated-word handling, per-doc dedupe, registry presence.
Retrieval-side only — synonym edges reshape FTS5 candidate selection
via synonym_expand at query time, never enter audit_mode / cache_key
/ audit_event_hash. No governance hash bump, no cache invalidation.
Closes#000050 §2a's CPU/GPU abbreviation rows *upstream* of vec;
the Orwell-shape conceptual-allusion row remains the genuine #000050
justification. Operational follow-up (not code): run on each shard
via `arborist concepts derive --extractor acronym_parens` (CLI
surface itself is aspirational in docstrings; extractors are called
programmatically today). Next ID 000054 -> 000055.
Built bench/fixtures/5f/recombination-realcontext-v1.jsonl: 25 hand-built
recombination claims (deliberate re-attributions within real bench-qa
30 KB Wikipedia haystacks the pipeline produced — Mount Kenya into a
Kilimanjaro answer, Galileo onto all 63 Jupiter moons, Joshua Reynolds
onto the Mona Lisa, etc.). Ran the full 7-model × full-grid sweep vs
the 808 pooled STRICT. Result:
- bart-large-mnli / k=12 / max / θc=0.999 → 12/25 = 0.48 catch · 0/808 FP
(the real-haystack fp=0 leader)
- deberta-large-mnli / k=3 / margin / θc=0.995 → 6/25 = 0.24 (§7 #26's
'settled' config — 28/28 synthetic, 0.24 real-haystack: 4× over-estimate)
- roberta-large 0.12, MiniLM 0.08, deberta-base 0.04
So §7 #26's 'boundary closed' walks back to 'boundary PARTIALLY closed'
on real haystacks. The bottleneck is architectural: top-k by token
overlap misses the contradicting clause when it shares few subject-area
tokens with the answer (e.g. the Mount Kenya clause only shares 'Kenya'
with a Kilimanjaro claim — ranked low, NLI never sees it). Threshold
tuning doesn't lift the ceiling; a SEMANTIC candidate selector
(vec-driven, sibling of #000050/#000051's hybrid retrieval) does.
bart's pareto above fp=0: fp=0.011 catch=0.52, fp=0.057 catch=0.84,
fp=0.068 catch=0.92 — permissive operating points are on the menu if
fox+dav1d sign off. recommended_operating_point updated to
bart-large-mnli/k=12/max/θc=0.999; deberta-large/margin kept as the
synthetic-eval reference. Sixth meta-lesson instance: clean synthetic
eval doesn't predict bench-qa precision OR recall — neither contrived
dataset axis is load-bearing, only the real pipeline shape is.
Production verifier unchanged; falsification-hard stays 10/12. Still
SHADOW; runtime promotion fox+dav1d-decides.
(a) Mined the pooled n=1+3+5 bench-qa runs (808 distinct STRICT answers)
for natural recombinations at lowered θc≥0.7 → 37 would-fires, ALL
token-collision FPs on inspection (Mount Kenya pulled into a Kilimanjaro
answer, Dalí into da Vinci, Donovan into Superman). ZERO genuine
recombination errors — the boundary is theoretical-in-practice; the
failure mode is the candidate selector (top-k by token overlap) pulling
different-entity same-subject-area clauses.
(b) Re-ran the grid against the 808-cell pooled STRICT set: the §7 #25
'max@0.96' was itself a small-sample artifact — the n=5 444-cell set
lacked the high-confidence spurious hits the pooled set has. On 808
cells θc goes back to ~0.995, and at θc=0.995 only agg=margin still
catches 28/28 (max gets 27/28). microsoft/deberta-large-mnli / k=3 /
agg=margin / θc=0.995 → 28/28 synthetic recombinations · 0/808 pooled
real STRICT FP · 0/26 synthetic legit — the ONLY config in the
7-model×full-grid sweep that hits 1.0/0.0 on 808 cells, held at n=3
too. recommended_operating_point reverted to margin@0.995.
Realistic next check: ~20-30 hand-built synthetic-recombination-vs-
real-bench-qa-context fixtures (real haystack, deliberate re-attribution).
Still SHADOW; runtime promotion fox+dav1d-decides. Production verifier
unchanged; falsification-hard stays 10/12.
Meta-lesson instance five: a bigger sample can vindicate a config a
smaller one made look unnecessary — re-confirm the config choice (not
just the threshold) each time the denominator grows.
Enumerates the concrete query-words-share-zero-tokens-with-target-title
cases the §2 gate's "semantic-allusion fixtures" must include, as a
running list: Orwell→Eastasia (genuine conceptual allusion — the case
that justifies the vec layer), "what is a CPU?"→Central processing
unit, "what is a GPU?"→Graphics processing unit (abbreviation→expansion
subclass — also fixable upstream by a concepts/ synonym edge; bench
records which fix closes each row). Field cases 2026-05-13, fox.
#000053 fixed the verifier's separate acronym blind spot but not this
retrieval gap.
ARBORIST_NLI_SHADOW=1 make bench-qa BENCH_QA_N=5 → 1125 cells, 444 real
STRICT (5x n=1, 1.6x n=3). Re-ran the 7-model mega-grid: the lexical-
candidate NLI veto robustly clears the §7 #12 gate with
microsoft/deberta-large-mnli / k=2 / agg=max / guard=max_entail /
θc≈0.96 / θe=0.9 → 28/28 synthetic recombinations (incl. both fixtures)
· 0/444 real STRICT FP · 0/26 synthetic legit FP, ~4 pts θc headroom;
roberta-large-mnli equally good (k=2/max/θc=0.95). Resolved: agg=max +
max_entail guard is the robust score-shape (§7 #24's margin win was a
sample tie); the large checkpoints (~350-400M deberta-large-mnli /
roberta-large-mnli) hit 1.0/0.0, bart-large (similar size) only ~0.71,
deberta-base at the cliff (27/28 here, 11/28 at n=3), small models
(MiniLM-82M, deberta-v3-small) cap at ~0.82 — so §7 #18's 'MiniLM is
the cost-pick' is OVERTURNED by the proper-n evidence; k=2 consistent
winner; int8-ONNX costs ~1 catch. recommended_operating_point updated.
Remaining: a bench-qa-derived recombination set (the one check not
done); n=9 if dav1d wants more; still SHADOW; runtime promotion is
fox+dav1d-decides. Production verifier unchanged; falsification-hard
stays 10/12. This run is the worked example behind CLAUDE.md's new
bench-maxing line.
`arborist.qa.evidence._content_tokens` dropped every token under 4
chars, so a short all-caps acronym (CPU, GPU, DNA, FBI, USB…) never
registered as a content token — which defeated Rule 8
(_claim_title_overlap / TITLE_MISMATCH), the subject-tokens-absent
check (Rule 9), the bare-name-claim guard, and spotlight-excerpt token
selection whenever a question/claim's topic IS an acronym. The field
case: `what is a CPU?` cited to the "CPU design" article tripped
TITLE_MISMATCH even though claim and title both contain "CPU".
Fix: keep a token if it's an all-caps 2-3-char alpha run in the source
text; everything else unchanged. The change only ever ADDS tokens, so
TITLE_MISMATCH / SUBJECT_TOKENS_ABSENT / BARE_NAME_CLAIM can only stop
firing, never start — monotone toward fewer spurious demotes; no
STRICT→non-STRICT transition is possible from it.
Versioned: `content_token_rules: "v2-acronym-aware"` added to
runner.DEFAULT_POLICY + query.DEFAULT_QUERY_POLICY +
keys._VERIFIER_POLICY_FIELDS → folds into verifier_policy_hash, prior
cache records orphan on lookup (by design; same discipline as
base_version / hyphen_fold_v1). Does NOT touch the retrieval
abbreviation→expansion gap (CPU→Central processing unit — #000050 vec
hybrid / concepts/ synonym edges; the root cause of the satellite
retrieval). 8 new tests; full suite green (2502); bench-qa-smoke clean.
Next ID 000053 -> 000054.
ARBORIST_NLI_SHADOW=1 make bench-qa BENCH_QA_N=3 → 275 real STRICT cells
(3x the n=1 sample). Re-ran the expanded grid (7 aggregations incl.
margin = max_clause(p_contra - p_entail), paired-entail guard variant,
θc to 0.999, --extra-models) over all manifest models + 4 extra
xsmall→large (microsoft/deberta-large-mnli, roberta-large-mnli,
nli-deberta-v3-small, deberta-v3-xsmall), synth-28 recombination vs the
275 STRICT cells. Result: the §7 #23 deberta-base/k=2/θc=0.99 config
does NOT survive — it catches only 11/28 at θc=0.995 (which the larger
STRICT sample forces). BUT the broader sweep found the config that does:
microsoft/deberta-large-mnli / k=3 / agg=margin / θc=0.995 → 28/28
synthetic recombinations (incl. both fixtures) + 8/12 falsification-hard,
0/275 real STRICT FP, 0/26 synthetic legit FP — a passing config at
proper n. Findings: margin is the right score-shape (single threshold,
folds the guard in); the specific checkpoint matters more than param
count (deberta-large-mnli wins clean, deberta-base collapses,
roberta/bart ~0.71-0.75 — no 'bigger is better' law). recommended_operating_point
updated. Still SHADOW; runtime promotion needs a bigger STRICT sample +
a bigger recombination set + fox/dav1d sign-off. Production verifier
unchanged; falsification-hard stays 10/12.
Meta-lesson sharpened twice: clean eval ≠ bench-qa precision (§7 #18→#20),
default config ≠ best config (§7 #22→#23), small FP sample ≠ large FP
rate (§7 #23→#24).
