# Teaching-brief synthesis: what the glitches teach, and where it points Synthesis of 42 raw recipes across 6 lenses (frames-order, metadata-reading, pairing-symmetry, number-systems, solved-pattern, application-map), cross-checked against `theories/DECISIONS.md` (through D42) and the brief's DEAD list. --- ## 1. Meta-pattern: what the glitches consistently teach, and the one target they point at Read across all six lenses, the recipes converge on the same lesson from a dozen different directions: **The word is never IN the piece's own essence/number fields — a correct value or index points at the WRONG LIST, the WRONG COLUMN, or the WRONG DIRECTION, and the fix is indirection, not translation.** This is exactly podGALLERY's solved grammar ("right index, WRONG LIST: SLIP-39 -> BIP-39") and it recurs as: - **Wrong list, same index** — reading BIP-39 back-to-front instead of front-to-back (metadata-reading); reading a painted number into SLIP-39 instead of BIP-39 (pairing-symmetry, application-map); reading the same digit string in a different base (number-systems, ×3 independent recipes on #465/#466, corrupt-11 skull/robo members, and painted marks). - **Wrong piece, same rank** — a confirmed 11-piece selection's mint-order RANK is correct, but it points not at those 11 pieces' own fields, but at whichever pieces sit at ranks 1–11 of a *different* ordered list entirely (title-alphabetical) — the solved-pattern lens's index-transplant recipes, run twice (on the corrupt-11 and on the prime-rule set). - **Wrong column, same piece** — Type and SubType can trade places (#62); a key itself can be misspelled (#383); absence can be structural for a whole drop (spooky Essence) — teaching "check every field, not the one you assume," which several lenses turned into a deck-WIDE census (rarest Type/SubType globally; deck-wide one-letter-off typo scan) rather than a read confined to an already-known eleven. - **Pairing/duplication is licensed, not noise** — echoed by deliberate pairs, TWIN vs ENTANGLED, and the three-way Rainbow Rosetta stone — repeated across four lenses as a standing instruction to treat any duplicate as a signal, not an error to fix. A secondary, less-explored thread: several structural glitches (backward-Fib splice gaps, held frames, palindrome mirroring) may encode the **phrase's SLOT ORDER**, not its word CONTENT — an axis almost nothing in the existing DECISIONS.md history has tested, because every past sweep asked "which 11 pieces, read how" rather than "which of the 12 slots does this pin down." **Bottom line: the single most consistent TARGET across lenses is the BIP-39 wordlist itself, approached through indirection (reversed, wrong-list, rank-transplanted, or re-based) — not the pieces' essence/number fields, which D38b already closed exhaustively.** --- ## 2. Dedupe notes (credited across lenses) - **Eyes-as-binary (H-2)** — raised independently by pairing-symmetry and number-systems (twice). Same still-open, memory-confirmed hypothesis; not yet built (needs eye-position extraction). One entry in Tier C, credited to all three. - **Painted-mark re-basing** (SLIP-39 lookup, hex reread, decimal-point word-halves) — raised by pairing-symmetry (×2) and number-systems (×1). Same 3–4 numeric marks, different lookup tables tried on each; capped at ≤4 members regardless of base/list. Collapsed into one Tier C entry. - **HiddenType misfile re-verification** — metadata-reading and number-systems both re-ran the label-vs-true-class check across the whole deck; both land on the same already-known 4 (#32,#119,#161,#187), confirming closure rather than adding a 5th. Collapsed into one hygiene note. - **Type-else-essence fallback register** — the SAME mechanical trick (read Type if it's a BIP-39 word, else fall back to Essence) is applied by application-map to three different selections: Silver-11 (already dead, D34 A3/A4), TWIN-11 (essence already dead on this exact selection, D34 B1), and the newly-found Ultra-11. One method, three targets; only the Ultra-11 application is genuinely new, and it inherits the same degenerate texture (6–8 distinct words of 11, heavy essence-fallback) as its two dead siblings — treated as a weak sibling, not three separate leads. - **Palindrome-word wordlist filter** (words spelled the same forwards/backwards: dad, eye, level, mom, radar) — computed identically by pairing-symmetry and application-map. Collapsed into one Tier C entry. - **Index-transplant grammar** — solved-pattern applies the same "rank in list A -> read list B at that rank" mechanic to two different source lists (corrupt-11, prime-not-Fibonacci set). Same method, two applications — kept as two entries (A1 computed now, B2 needs a build) because they use different source populations. --- ## 3. Cross-check vs the DEAD list — drops, demotions, and fresh angles on dead ground **Dropped outright (re-runs of exhausted registers, no fresh angle):** - Drop-size/wordlist-size divisibility speculation (number-systems) — self-identified as the same shape as the already-DEAD base-32 block-partition work (D32/D33); no new pairing field changes the conclusion. - Backwards-Fib position-sum/frame-count geometry as a direct word index — already dead per D27/D28 (gap sequence is identical per piece, non-discriminating). **Demoted (produce a literal word set, but the register is a near-miss of something already dead, or degenerate on inspection):** - TWIN(11)/Ultra(11) Type-else-essence fallback (application-map) — essence was already swept dead on the TWIN-11 selection itself (D34 B1); this hybrid is 6–8/11 essence in disguise and comes back with heavy word repetition on both applications. Kept in Tier A as "computed," demoted out of the top 5. - Character-offset-of-first-mismatch index (solved-pattern) — a fresh mechanic in principle, but the pass that computed it used the *display Title* field, which mostly just measures where a "(16/128)"-style mint suffix begins, not the real glitch. Needs the raw on-chain Type/SubType text recomputed before it's trustworthy; also needs a D42-style check that one consistent character-position definition works across all 11 corruption verbs before sweeping. Demoted to Tier B. - Backwards-Fib stride-from-a-pinned-anchor (solved-pattern) — the anchor itself has to be guessed (word #1, GAN#0's essence, HODL's first word), which is exactly what D36's no-guess doctrine forbids unless the anchor can be pinned independently first. Kept as Tier C, flagged explicitly as guess-dependent. **Fresh angles on dead or near-dead ground (worth flagging, not dropping):** - Reversed-wordlist read on the fib11 class's already-decoded N-values (metadata-reading) — sits close to the banned "word-transforms (reversal/atbash/Caesar) as a family," but that ban is about flipping the letters *inside* a single already-found word; this reverses the *order of the list itself* before indexing — a different operation. Flagged for a 5-minute doctrine check before spending sweep budget, per the recipe's own caution. - Index-transplant onto title-alphabetical rank (solved-pattern, ×2: corrupt-11 and prime-rule sources) — essence register is DEAD "on every ratified eleven," but here essence is read on a *different, newly-derived* set of 11 pieces (the alphabetical-rank matches), which have never been a "ratified eleven" selection before. Mechanically distinct from the banned register even though the reading field is the same one that's died everywhere else — tempered enthusiasm accordingly, but genuinely unswept per the registry audit. - Deck-wide one-letter-off typo census (metadata-reading) — extends the already-ratified 9-pair "deliberate pairs" class with 2 new pairs, reaching exactly 11 pairs total. Not a re-run: nobody had scanned the *whole* deck for this specific typo family before; only the 9 pairs already on file had been hand-found. - Global Type/SubType rarity-pool census (metadata-reading) — reuses the SWAP lesson (#62) as a population-wide filter instead of a single-piece read; distinct from every prior essence/number register because the selection criterion (cross-column rarity) has not been run before. **Confirmed negatives worth recording so they aren't re-proposed** (closed by this pass, cited briefly, no further sweep needed): HiddenType label-vs-true-class census (exactly the same 4, no 5th); misspelled on-chain KEY census (exactly 1, isolated — #383); Essence missing as a whole-drop on-chain trait for spooky (structural, confirms the drop's own payload lives in its octal numeral instead); Prime-not-Fibonacci rule tested against spooky/ ghost/solstice native numbering (clean falsification on all three — closes D27's "unverified" caveat as answered-negative); Name-vs-GANg mismatch count (exactly 10, not 11); adjacent- letter-swap wordlist filter (exactly 6 pairs, short of 11); DELETE+field-shift on the ENTANGLED gap pieces (3 of 5 confirmed dead-end, 2 of 5 still genuinely unexplained). --- ## 4. Tiers ### Tier A — sweepable 11-word sets, computable NOW from a non-essence-only, objective target **A1. Index-transplant: corrupt-11 mint-rank → title-alphabetical rank → Essence** *(solved-pattern lens)* — HIGH promise, genuinely unswept per prior registry audit. Mint order 1–11 of the Pindar-confirmed corrupt-eleven {16,17,44,48,58,62,66,81,173,281,309} maps onto ranks 1–11 of all 512 bitGANs sorted alphabetically by title; read each landed piece's Essence. > **GOAT · FOX · BULLET · CROSS · AIR · GRAB · GODDESS · BLIND · EIGHT · DIVIDE · BEAN** (from #194, #320, #397, #347, #44, #204, #195, #131, #137, #449, #21 — all 11 confirmed present in bip39.txt.) **A2. Reversed-wordlist read of the fib11 class's already-decoded N-values** *(metadata-reading lens)* — HIGH promise, but flag for a doctrine check first (see §3). Take the fib11 class's 11 binary-suffix N-values (19,23,29,31,37,41,43,47,53,61,287 — piece order 148,152,158,160,166,170,172,176,182,190,287); instead of indexing N words in from the front of bip39.txt, index N words in from the **end**. > **WORK · WONDER · WISDOM · WINTER · WIN · WIDTH · WHISPER · WHEEL · WELCOME · WEAPON · SWING** (10 of 11 start with "w" — real authored-feeling texture; "swing" is the one seam piece, #287, whose title carries a literal decimal number instead of binary.) **A3. TWIN(11 via EC40) — Type-if-BIP39-else-Essence fallback** *(application-map lens)* — DEMOTED: mostly essence-fallback (8/11 slots), and essence already swept dead on this exact selection (D34 B1). Mint order 47,59,73,121,124,132,139, 323,373,379,395. > alarm · also · animal · smoke · ghost · brother · echo · ghost · exit · monster · ghost **A4. Ultra-11 (SubType='Ultra', new census) — same fallback method** *(application-map lens)* — fresh SELECTION, recycled register; treat as a weak sibling of A3, not independent evidence. Mint order 313,314,315,332,361,362,377,393,412,426,430. > example · monster · space · spider · ghost · skull · monster · ghost · monster · upper · dynamic ### Tier B — needs a build (a rule, a threshold, or a raw-field recompute) before it's sweepable - **B1. Splice-mask / Fibonacci-mirror slot scaffold** *(frames-order lens, 3 chained recipes)* — internally consistent arithmetic chain: the fib11 gap sequence 5,3,2,1,1 sums to 12 and walks a 12-slot cycle to hit slots {5,8,10,11,12}; mirroring each (13−i) gives {1,2,3,5,8} — exactly the Fibonacci "protected" positions already documented elsewhere in the project, an independent cross-confirmation from a completely different glitch class. Leftover "free" slots: {4,6,7,9}. Held frames (#479→slot 3, #480→slot 12) independently land on two of these same slots. This is a **phrase-ORDER** hypothesis, not a word source — needs word content plugged into the slots (e.g. from A1/A2) to become a full test. - **B2. Prime-not-Fibonacci set → index-transplant onto title-alphabetical rank** *(solved-pattern lens)* — same solved mechanic as A1, unbuilt: needs the source-list frozen (first 11 by mint, vs. one representative per drop) before it can be computed to literal words. - **B3. Deck-wide typo-pair census → 11 pairs total** *(metadata-reading lens)* — 9 ratified + {9,73} + {170,175} = 11 pairs; needs one tie-break rule for #175 (which sits in two of the eleven pairs) before a clean one-representative-per-pair selection exists to sweep. - **B4. Global Type/SubType rarity-pool census** *(metadata-reading lens)* — 'Mechanic' appears exactly once deck-wide; 'Alien' and 'Quantum' twice each; needs the rarity threshold widened methodically to see whether the union of low-count members reaches exactly 11. - **B5. Character-offset-of-first-mismatch, recomputed on RAW on-chain Type/SubType** *(solved-pattern lens)* — self-flagged unreliable in its display-Title form (mostly measures mint-suffix boundaries); needs the raw field text and a D42-style single-logic check across all 11 corruption verbs. ### Tier C — needs the collector's eye, a real detector build, or is a capped small component - **C1. Eyes-as-binary (H-2)** — confirmed still-open in project memory across 3 lenses; blocked on eye-position extraction (needs Michael's eye or a detector), not on sweep budget. - **C2. Painted-mark re-basing** (SLIP-39 lookup, hex reread, decimal-point word-halves) — capped at ≤4 numeric marks project-wide, can't reach 11 alone. One loose thread worth a cheap look: concatenating #165+#207's marks (81+4→"814") lands on bip39[814]="grant" — literally #207's own essence word, an unexplained coincidence. - **C3. Palindrome-word wordlist filter** — dad, eye, level, mom, radar; objective and computed, but only 5 words with 1 real-world echo (#140 essence "eye"). - **C4. Held-frame direct bip39 index lookup** (27→"alter", 24→"ago") — 2-word component only. - **C5. TRUE/FALSE (#363/#473) indexing tie-break** — first check shows no clean parity split; park unless revisited. - **C6. Backwards-Fib stride-from-a-pinned-anchor** — anchor choice is itself a guess, against D36 doctrine; park unless an anchor can be pinned independently first. --- ## 5. Top 5 recommendations 1. **A1 — Index-transplant (corrupt-11 → title-alphabetical rank → Essence).** Run GOAT FOX BULLET CROSS AIR GRAB GODDESS BLIND EIGHT DIVIDE BEAN (+ its reverse) through the plant-check/checksum pipeline against `0x18f87ec9c527aba1db44f715456bf28b0dae478d`. 2. **A2 — Reversed-wordlist read of the fib11 N-values.** Five-minute doctrine check first (confirm this reverses list-order, not word-letters, so it's distinct from the dead reversal family); then plant-check WORK WONDER WISDOM WINTER WIN WIDTH WHISPER WHEEL WELCOME WEAPON SWING (+ reverse). 3. **B2 — Prime-not-Fibonacci → index-transplant.** Freeze the source list (first 11 by mint vs. one-per-drop), compute both against title-alphabetical rank, plant-check each. 4. **B1 — Splice-mask/mirror slot scaffold.** Take the phrase's SLOT STRUCTURE (hit {5,8,10,11,12} / Fibonacci-protected {1,2,3,5,8} / free {4,6,7,9}) and test it as an ordering key against the word sets from A1/A2, to see if it beats plain mint order on the checksum. 5. **B3 — Deck-wide typo-pair census (11 pairs).** Pick a tie-break rule for #175 (member of two pairs), assemble the one-representative-per-pair 11-set, plant-check.