THE MOVE CATALOG — the teacher's compile moves, registered

Standing law (owner-ruled 2026-09-02). The owner's words: "I like the catalogue registry. This is probably going to be one of the more important records to keep because we are in discovering deep logic... it needs to be updated when we find this logic." And on adoption: "Yes, make this change. This is very exciting. We are finding deep logic now."

This file is the canonical registry of TEACHER MOVES — the recorded reasoning forms by which the Mishnah/Talmud layer compiles the 24-book code: fills the gaps the bare ink leaves open, and gets from the text's mechanisms to the answer key's tables. It is updated THE SITTING a new move (or a new exemplar of a known move) is discovered — a standing duty, no owner approval per update; every machine-administered update is labeled delegated, and the owner can overrule any entry. Entries are never deleted; corrections append.

Relation to the other registries: logic/MIDDOT.md holds the tradition's own numbered inference rules (13 law / 32 narrative) — the tradition's names for its logic. THIS catalog holds the moves as the COMPILER meets them: operational forms, each with measured exemplars, machine-replayed where a cold run exists. Where a move IS one of the middot, the entry says so. World/step9/RULE_CATALOG.md holds the compiled output rules (R-numbers); this file holds the moves that produce them.

Format per entry: the move, in plain words; what gap it fills; the recorded exemplars (each with its source rows and, where run, the cold-run cell it filled); middah correspondence if any; date found.

M-01 — THE DIFF (recover a hidden parameter from branch outputs)

The move: two near-identical code passages never state what distinguishes them; feed both the same input, compare the outputs, and infer the hidden configuration from which one is stricter — anchored by the principle that liability tracks benefit. Fills: hidden parameters the ink never names. Exemplars: the paid/unpaid keeper — both paragraphs say only "to keep"; on theft, paragraph one exempts by oath, paragraph two pays; stricter = compensated (Babylonian Talmud Bava Metzia 94b:7-10, with the double-payment objection raised and answered). Cold-run: cold_run_guardians.py, role-assignment step, 12/12. Middah: none of the thirteen — this is SEVARA, reasoned analysis, which the Talmud holds as a source in its own right ("it stands to reason," and elsewhere: why do I need a verse? it is reasoning). Found: 2026-09-02 (the code hunt).

M-02 — THE A-FORTIORI (generate a missing cell from a stronger one)

The move: if the rule holds in the harder case, it holds all the more in the easier one — the tradition's own qal va-chomer (light- and-weighty), first of the thirteen middot. Fills: cells absent from the ink entirely. Exemplars: LOSS never appears in the keepers' nine verses; "if theft, which is near to circumstances beyond control, pays — loss, near to negligence, all the more so" (Bava Metzia 94b:15, "they say in the West"). Cold-run: the paid keeper's loss cell. Middah: #1, QAL VA-CHOMER (light-and-weighty) — an EXACT match: this move IS the first of the thirteen, machine-replayed. Found: 2026-09-02.

M-03 — THE COMPARISON (carry a rule between passages by their shared frame)

The move: two passages about the same institution share structure; what one states explicitly is carried to the other through the shared frame ("just as below... so too here"). Fills: rules stated in one variant but needed in a sibling. Exemplars: the borrower's theft/loss liability, carried from the paid keeper's passage (Bava Metzia 94b:19), sealed by all-benefit-is-his (94b:10). Cold-run: borrower theft/loss cells. Middah: the #2/#3 family — the "just as below" frame-comparison works like the founding-case and verbal-analogy forms; the gemara runs it as structural comparison rather than naming a middah. Found: 2026-09-02.

M-04 — THE ROUTING (classify an unnamed input onto an existing branch)

The move: a case type the code never names arrives; the teacher files it under an existing branch's rules — and logs the recorded dispute over which branch. Fills: the gap between the code's N variants and the world's N+1 case types. Exemplars: the RENTER — the code defines three keepers; the answer key lists four; the renter follows the paid keeper's row (Mishnah Bava Metzia 7:8; the four-fold division affirmed and the routing disputed at Bava Metzia 93a:17), hooked on the ink's own hire-clause ending Exodus 22:14. Cold-run: the renter row, 3 cells. Middah: none named — this is the classification practice the Talmud performs when sorting derivatives under primaries (the avot/toladot sorting that opens Bava Kamma). Found: 2026-09-02.

M-05 — THE THRESHOLD-PARSE (read a numeric constant out of the verse's own tokens)

The move: a state-transition constant is encoded in the verse's wording; the teacher tokenizes the phrase and counts — and preserves rival tokenizations by name. Fills: numeric thresholds the code states as idiom. Exemplars: the goring ox's "from yesterday and the day before, and its owner has not secured it" → THREE gorings flip innocuous (תָּם "innocent") to forewarned (מוּעָד) — Abaye's count and Rava's count both preserved (Bava Kamma 23b:17-18); the counter's SEMANTICS debated at 24a:9 (about the ox, or warning the owner?). Cold-run: the threshold cell. Middah: not of the thirteen — this is the REDUNDANCY EXPOSITION of the Akiva school (every seemingly extra word load-bearing); the Abaye/Rava split is literally a dispute over whether the doubled idiom is expounded word-by-word — a recorded parser-settings disagreement. Found: 2026-09-02.

M-06 — THE GENERALIZATION (one ink instance becomes the class rule)

The move: the code states an output rule inside one case; the teacher generalizes it across the sibling classes that share the interface. Fills: common-output rules stated once. Exemplars: "the best of his field and the best of his vineyard shall he pay" (Exodus 22:4, in the grazing case) → damages are paid from the best, across all four primary categories (Mishnah Bava Kamma 1:1's common-denominator clause; the discussion at Bava Kamma 6b:11). Cold-run: the best-of-land cell. Middah: #3, BINYAN AV (the founding case) — one stated instance founds the rule for the class sharing its feature; the Mishnah's own "common denominator" is this middah's vocabulary. Found: 2026-09-02.

M-07 — THE CROSS-MODULE IMPORT (a rule from another book reaches in)

The move: an event or constant declared in a different book of the code overrides or completes the local passage; the teacher documents the wiring — including that the code "needed to write" the connection. Fills: local gaps whose answer lives in another module. Exemplars: (a) the JUBILEE (Leviticus 25) releases the Hebrew slave — even the pierced one whose ink says "forever" (Mishnah Kiddushin 1:2; Kiddushin 15a:19). (b) HUMILIATION, the fifth indemnity, sourced from Deuteronomy 25:11-12. (c) THE CALL-SITE DEPENDENCY, the pass's largest find: eye-for-eye-as-MONEY is derived through Leviticus 24:21-22 — the formula's only other exact occurrence in the canon (machine scan: Exodus 21:24 + Leviticus 24:20 and nowhere else) — so the Exodus seat cannot be finished without the Leviticus call (Bava Kamma 83b:10, 84a:1). Cold-run: Jubilee, humiliation, and damage-as-money cells. Middah: mixed — the call-site derivation runs on #2 (GEZERAH SHAVAH, the verbal analogy on the shared striking-word) with the harmonization family (#13, two passages resolved together) behind it; the Jubilee wiring is the Talmud's NECESSITY ANALYSIS ("it was needed to write both"), practice rather than a numbered rule. Found: 2026-09-02.

M-08 — THE EXEMPLAR-GENERALIZATION (named instances imply the category)

The move: the code names two concrete instances; the teacher asks "granted, these are WRITTEN — what of the rest?" and derives the category they exemplify; the answer key then enumerates the category's members. Fills: the gap between the ink's examples and the law's class. Exemplars: the freed slave's EYE and TOOTH (Exodus 21:26-27) → the class "limb-tips that do not regenerate" (Kiddushin 24a:6) → twenty-four members enumerated (Mishnah Negaim 6:7). Cold-run: the class cell, pass 2. Middah: #3, BINYAN AV FROM TWO VERSES — the thirteen explicitly distinguish the founding-case from ONE verse and from TWO; eye and tooth are the two-verse form (the common-feature argument). Found: 2026-09-02.

M-09 — THE KEYWORD DICTIONARY (the language manual for the code's own words)

The move: the teacher maintains corpus-wide semantic entries for the code's operator words — the senses of a keyword, and the enumerated exceptions to its default. Fills: disambiguation the parser needs before any case can be cut correctly. Exemplars: (a) כִּי ("when/if") "has four distinct meanings: if, perhaps, rather, because" — Reish Lakish (Gittin 90a:10). (b) "Every 'if' (אִם) in the Torah connotes optionality, EXCEPT FOR THREE" — Rabbi Yishmael, with the three enumerated (Mekhilta DeRabbi Yishmael, Tractate Bachodesh 11:11). Used by: the verse-initial-KI parse rule in both cold runners. Middah: none — this is the LEXICON, the layer beneath the middot: the thirteen presuppose that the keywords are already correctly read, and these entries are the tradition doing that reading corpus-wide. Found: 2026-09-02.

M-10 — THE POINTER-FETCH (a label in one book, the constant in another)

The move: the code states a value by NAME rather than number; the named constant is defined in a different module; the teacher records the link that resolves the pointer. Fills: unpriced fines, unspecified measures. Exemplars: the seducer pays "money like the dowry of the virgins" (Exodus 22:16) — a label, no number; the constant, fifty silver, lives in Deuteronomy 22:29; the link recorded (Ketubot 29b:3 parsing the triple mention; the payments table at Mishnah Ketubot 3:4). Cold-run: the FETCH-50 cell, pass 2. Distinct from M-07: the import brings a RULE in; the fetch resolves a NAMED VALUE. Middah: #2-family — the recorded resolution runs through the verbal-analogy machinery on the shared dowry/virgin ink (with the Akiva-school triple-mention inclusion at Ketubot 29b beside it). Found: 2026-09-02.

RUNNING COUNT: 10 moves, all exemplified at pulled rows, 9 of 10 exercised by a cold run (M-09 is used by the parser itself).

THE MIDDAH CORRESPONDENCE, measured (2026-09-02, owner's question): roughly half the moves ARE numbered middot — M-02 is #1 exactly (qal va-chomer); M-06 and M-08 are #3 (binyan av, one-verse and two-verse forms); M-07 and M-10 run on #2 (gezerah shavah) with the #13 harmonization family behind the call-site. The other half is the Talmud's DOCUMENTED PRACTICE beyond the numbered list: sevara (M-01), classification (M-04), the Akiva-school redundancy parser (M-05), necessity analysis (inside M-07), and the lexicon (M-09). Reading: the thirteen middot are the PUBLISHED instruction set — the formally licensed derivation forms — while this catalog records the FULL observed instruction set in use, which is larger. The published subset is the natural first target for middot-as-data. The compile-dependency graph so far: Exodus 21 ← Leviticus 24 (M-07c); Exodus 22 ← Deuteronomy 22 (M-10); Exodus 21 ← Leviticus 25 (M-07a); Exodus 21 ← Deuteronomy 25 (M-07b).