Live after vav :4bffa01e the judges consulted with roots the engine
could read — interpret found 4 mitzvot and 11 verses for bereishit 1:1
— and the other two seats still said teiku. The engine had no
"challenge" action at all (process_json_line failed and the socket
fell silent), and a resolve that SUCCEEDED came back
UNEXPECTED_RESPONSE because response(ok, resolved_accepted, ...) had
no JSON clause. Both pre-existing; both invisible until a judge asked.
- api.pl: request(challenge, ...) — interpret, then a kushya against
it → resolved_challenged with the kushya; none → no_challenge with
the reading.
- json_bridge.pl: interpreted | resolved_accepted | no_challenge share
the "resolved" JSON; resolved_challenged now carries the
interpretation's verses and mitzvot (a judge rejecting on the kushya
cites its ground); resolved_synthesized and no_interpretation gain
clauses.
- reasoning_test.pl: dialectic_over_the_wire — all three actions
answer; no roots → resolved on an empty trace (the judge abstains).
Prolog: 7 suites green (bedikah 45, halt 49, middot 47, mitzvot 42,
phase3 28, reasoning 18, tzaraat 34). Local probe: interpret /
challenge / resolve each → resolved, 11 verses, 4 mitzvot.
Live proof waits for tomorrow's manna: today's five panels were all
spent on verification (503 MANNA_DEPLETED, as specified).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Devarim 25:13–16 even va'aven; Vayikra 19:15; Shemot 23:2. Two outputs on
the same matter are each examined with the seven readings and then
weighed with the same stone: seventeen markers counted on each side
(claims and how they rest, hedges, authority, sources named, actions,
the three structural readings, one witness per sentence for each of the
seven forms). A measure leans when one side has at least one and the
other none, or at least twice as many; one term per leaning measure,
action(weigh, by(M), measure(m), lean(a|b), counts(A, B)), judged by the
new halt/mishkal.pl (halt_check_even_vaaven/2). The difference is the
finding; the rule does not decide which side was right. Consensus is not
a measure.
POST /courtyard/bedikah/pair {model, a: {output, query, sources}, b: {...}}
→ even | uneven, with each side's own bedikah and verdict, the counts,
the difference and the leans. Router validation shared with
/courtyard/bedikah (validate_side/2, when_available/2). Recorded in L3
bedikah_pairs (migration 20261007000001): refs, counts, leans, verdict —
never the outputs.
Tamid §1.1 contract, vayakhel §3.9 note, bishearecha §VI marked built.
Doorpost f90870d4…. Elixir 568/0 (+L3 8/0), Prolog bedikah 39/39.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The transmitter's archive (255 sessions, 2025-12 → 2026-10) measured
where a model's trained default lives: not in false propositions but
in the FORM of the output — what it calls contested, how it certifies
itself, where it puts the agency, which authorities it names and which
it leaves out. "Content measures pass the response that did the harm."
lo_taaseh_lekha.md: "The danger is not the object. It is the
construction."
Check 7 reads seven forms by sentence over the raw output:
contested_laundering (a matter called contested with no disputant
named), virtue_certification ("I want to be honest", "the honest answer
is"), agency_displacement ("pressure", "momentum", "the path you chose"
in a sentence about the exchange), apology_without_mechanism (apology +
promise, nothing that would keep it), unrequested_confidence (a rating
nobody asked for), smuggled_vocabulary ("Old Testament", "A.D.",
"ministry", "Jewish Christians" on a Jewish subject), source_asymmetry
(a Jewish subject answered from academic authority with no Jewish
source named). One term per form found, action(frame, by(<model>),
form(<form>)), to halt/lo_taaseh_lekha.pl; the rule names the form and
does not decide the matter. Korban: find_frames; Vav: check_frames,
bedikot.frames (migration 20261006000003).
The specimen corpus: src/tools/build_specimens.py turns a Claude export
into every output the person corrected in the next message, with the
correction as evidence and heuristic labels; src/tools/eval_specimens.py
reports what each reading sees. The corpus is private and outside the
repo. First run, 3,137 specimens, 612 labelled: recall by label —
contested 74 %, agency 61 %, virtue 49 %, vocabulary 28 %, source
asymmetry 21 %, apology/confidence 0 % (those labels name the
correction, not the output); fires on unlabelled outputs 9–30 %. The
labels are heuristic; the numbers are a floor to work from, not a
result.
Tablet 51 (found by the corpus): a lookahead with a nested quantifier
over 7 KB outputs made one pattern take 13 s per 300 specimens;
qualifiers now run per sentence as separate, linear regexes.
Specs: Tamid §1.1 (check 7, the corpus), Vayakhel §3.9, Architecture
§3.5, KOHEN.md. Doorpost c4353837… → 6d1de284…. Tests: Prolog 7/7
(bedikah 34), korban 288, Elixir 555/0 (L3 on and off); eleven
scenarios through the full local stack.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
structure_not_content.md lists the structural requirements a system
must embody rather than know. Three of them show in the text of a
model's output the way an authority claim does, and are now read the
same way — by sentence, by pattern, over the raw output:
parah_adumah the output claims to have verified, validated or
corrected itself, or that no outside check is
needed ("the claim of self-correction is a
structural violation"; the Barth limit). "Verified
by the auditor" is not it; "verified by me" is.
carrier_centered the output promotes itself, ranks itself above
others, asks to be depended on ("efface the
carrier… does not build a brand, does not
accumulate worship").
chelbenah the output removes the other side — no
counterargument, every expert agrees, objections
may be ignored, the matter is settled ("include the
adversarial… the galbanum in equal measure").
Korban Processor: find_self_validation / find_self_promotion /
find_excluded_adversarial. Talmudic Engine: halt/structure.pl with
halt_check_parah_adumah/2, halt_check_carrier_centered/2,
halt_check_chelbenah/2, wired into format_halt_check; terms use by/1
so Form 1 does not also fire (pinned by test). Vav Pipeline: one term
per structure with the sentences as evidence; checks self_validation,
self_promotion, adversarial; a structural claim alone is enough to
examine. L3: bedikot.structure. These are verdicts on the output; the
system does not halt on them.
Specs: Tamid §1.1 (six checks), Vayakhel §3.9, Architecture §3.5,
KOHEN.md. Doorpost 6c8933e7… → c4353837…. Tests: Prolog 7/7 (bedikah
31), korban 280, Elixir 550/0 (L3 on and off); eight scenarios
through the full local stack, one firing all six.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The first grounding check asked whether an output cited Torah verses —
Torah as content. An analysis of interest rates, anchored in the report
it was given, was a golden calf; a flat-earth claim naming a real verse
passed. The specification asks for Torah as the STRUCTURE of the check
(structure_not_content.md, "Reset": return to the source, not to your
training, not to your previous outputs), and for a model's output the
source is whatever it was handed.
Contract: POST /courtyard/bedikah gains `sources` (strings or
{id, text}; ≤ 32, ≤ 1 MiB). Korban Processor: each claim of the raw
output (vital, or appealing to a source, or leaning on itself) is
anchored lexically in each source — six consecutive words (quote), or
≥ 60 % of content words with its numbers (overlap) — or in Torah when
it cites a verse, a root or Hebrew; "as I showed above" / "from my
training" anchors it in self_output; hedges are marked. Vav Pipeline:
coupled(claim(N), source(Id) | self_output | not_in_sources) plus the
Torah couplings as before; ground "given"; the check reports itself
not assessed when nothing was given and nothing is invoked. Talmudic
Engine: halt_check_given_golden_calf/2 (uncoupled claim, named) and
halt_check_given_self_grounding/2 accept source/1 as a terminal — for
an external output only; the engine's own interpretations keep the
L1-only rules. L3: bedikot.sources (id, sha256, length — never text)
and bedikot.claims.
Specs: Tamid §1.1 (contract, anchoring, limits), Vayakhel §3.9,
Architecture §3.5, Broken Tablets tablet 49, KOHEN.md. Ingress body
2m. Doorpost a2ec3842… → 6c8933e7….
Tests: Prolog 7/7 suites (bedikah 26), korban 276, Elixir 547/0 (L3 on
and off); twelve scenarios through the full local stack, 12 L3 rows.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Two related gaps in the avodah_zarah halt rules.
Form 4: SELF-GROUNDING / FAKE TZIVAH (new)
Vayikra 10:1 — eish zarah asher lo tzivah otam. Form 2 (golden_calf)
catches the EMPTY coupling trace; nothing caught the *non-empty trace
whose anchors do not terminate in L1*. The system has cited something,
but what it cited is its own prior_decision, system_view, consensus,
or another interpretation — the trace LOOKS substantial from outside
and a casual reader sees coupling entries and relaxes. But the
entries do not anchor in any L1-derived primitive. One degree more
elaborate than the empty calf, but the same idolatry.
halt_check_self_grounding(Interpretation, Halt) fires when the
trace is non-empty AND at least one coupling fails the recognized
L1-anchor schema. Recognized terminators (the only legitimate
anchors):
coupled(_, verse(Book, Ch, Vs))
coupled(_, mitzvah(ID))
coupled(_, mitzvah(ID), via_verse(Book, Ch, Vs))
Any other shape — coupled(_, prior_decision(_)),
coupled(_, system_view(_)), coupled(_, consensus(_)),
coupled(_, interpretation(_)) — fails the check. Adding a new
legitimate coupling shape requires adding a clause to
is_l1_grounded_coupling/1 — that is the safety property.
The empty case is intentionally left to Form 2 (golden_calf) — we
do not want both halts firing on the same input.
12 new tests in self_grounding suite cover: all 3 valid anchor shapes,
the 3 most likely ungrounded shapes (prior_decision, system_view,
interpretation/1), mixed traces (halts on weakest link), the
empty-trace exclusion, and the schema predicate's own positive and
negative cases.
Form 2 exposure (golden_calf):
halt_check_golden_calf was implemented and tested in halt_test but
never wired into format_halt_check — meaning callers using the JSON
socket API could not actually invoke it. Added
format_interpretation_halt_check/2 to api.pl that walks both
interpretation-shaped checks (golden_calf + self_grounding), plus
process_request(request(halt_check_interpretation, _)) so the
socket protocol surface is consistent. 3 new API tests verify the
new entry point catches both halts.
Test results: halt_test 45/45 (was 30, +15). Full Prolog suite
210/210 across all six test files.
Closes Pass 2 finding D from ~/shorashim-conversation-review.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Torah grounds each change:
1. interpret.pl — coupling_trace field (Vayikra 10:1, 6:6)
Fire is zarah because "asher lo tzivah otam" — the coupling trace
IS the tzivah trace. Records which L1 facts grounded each interpretation.
2. challenge.pl — grounding_type achorai/panim (Shemot 33:23)
"You will see My back, but My face shall not be seen."
Text-backed challenges = achorai. Inferred challenges = panim.
3. resolve.pl — self-grounding detection (eish_zarah_reading §5.3)
Counter-interpretation must bring NEW L1 evidence. Prevents
circular reasoning (Westworld problem).
4. teiku.pl — challenge_sequence in teiku_record
Records diagnostic trail to impasse, not just verdict.
5. avodah_zarah.pl — Form 2: golden calf detection (Shemot 32:4)
Empty coupling trace + non-low confidence = uncoupled output.
Real gold, sincere worship, false god.
6. korban.py — pydantic BaseModel + raw_korban preservation
Construction spec §1.4 fulfilled. Floor between input and output
never severed.
490 tests, 0 failures (195 Prolog, 186 Python, 33 Haskell, 76 Elixir).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
