# Methods synthesis: 28 raw recipes -> deduped, ranked, sweep-ready Source: 4 method-family agents (trait-words, lore-words, number-absence, classifier-order) applying the UNTESTED methods list (lore_catalog_runs/METHODS_BRIEF.md) to the ratified glitch sets. Every Tier-A word and every selection-membership claim below was independently recomputed against `deck_512.json` / `bip39.txt` in this pass (not just trusted from the raw recipe) — see the verification notes under each entry. Field values quoted match the deck exactly unless flagged. --- ## TIER A — sweepable now (concrete 11-word sets, paste-ready) ### A1. `S2-W10` radix-decode — backwards-Fib-eleven(11) — **top pick** Selection: `148,152,158,160,166,170,172,176,182,190,287` (the ratified/caveated fib-eleven; mint order == GAN-id order for this set, verified: mint 230,234,240,242,248,252,254,260,268,277,395). Rule: parse the 6-bit binary suffix baked into each robo/crypto-GAN Title (e.g. `(101011)`) as base-2 -> decimal N, then index N into bip39.txt. #287 is the one seam: its title carries a literal decimal `#287`, not a binary string, so N=287 directly — flag this piece if the sweep wants a binary-only variant. Verified N values (recomputed from Title/M_BinHexConv, not just trusted): 19,23,29,31,37,41, 43,47,53,61,287. **0-indexed word set (mint order):** `act, actual, admit, advance, again, ahead, air, album, allow, amateur, catalog` **1-indexed alt (bip[N-1]):** `across, actress, adjust, adult, afraid, agree, aim, alarm, alley, always, cat` CAUTION (my own cross-check, not in the raw recipe): the brief's DEAD list bans "Number registers (piece#, mint#, **binary-position**) on fib-eleven," and D27's overnight sweep log says fib11 was hit with "essence + all number registers + geometry x 7 orderings." Whether that phrase already covers "binary-position decoded then used as a bip39 INDEX" (this recipe) vs. "binary-position used directly as a seed value" (what's explicitly named dead) is not resolvable from the logs I can find — `theories/_registry.json` is empty and no sweep script naming this exact operation turned up. Novelty: **partial-overlap, not clean-new** — recommend a cheap registry/log check before spending real sweep budget, but it is cheap enough (2 candidate strings) to just run. ### A2. `S2-W13`/`S2-W23` — "11 person/entity tributes" (new selection) Selection: deck flag `G: Named after person/entity` has 14 members; subtract the 3 mythic-founder anchors #1/#2/#3 (essences abandon/ability/able — themselves the first 3 BIP-39 words, a separate already-dead register) to get 11: `210,211,228,246,264,282,445, 446,447,448,482`. NOTE: excluding the founders is the recipe agent's own reasoning, not a collector ruling — this selection is **not yet ratified**, only data-grounded. Mint order (recomputed, and it is NOT GAN-id order — verified via TokenIndex_mintorder): `210(310), 211(311), 228(331), 246(349), 264(369), 282(389), 447(585), 448(586), 482(625), 445(660), 446(661)`. **Word set in mint order** (all 11 confirmed valid BIP-39 words by direct lookup): `champion, escape, potato, mask, acid, recycle, cloud, elder, test, prize, flower` Per-piece basis (recomputed Essence fields shown for contrast): | GAN | homage | word | basis | Essence (for contrast) | |---|---|---|---|---| | 210 | noohp (community solver) | champion | **external, unverified** | awake | | 211 | mazzone | escape | in-corpus quote ("excaped into...") | aware | | 228 | Kevin Abosch | potato | **external, unverified** ("Potato #345") | barely | | 246 | Hackatao | mask | **external, unverified** (masked-figure motif) | belt | | 264 | Killer Acid | acid | literal pseudonym, high confidence | blame | | 282 | Eclectic Method | recycle | in-corpus, cross-validated via #423 | boost | | 447 | Brinkman | cloud | **= its own Essence — not actually distinct** | cloud | | 448 | StellaBelle | elder | in-corpus (theories/better_sentences.md) | legend | | 482 | Testy McTest | test | obvious placeholder reading | call | | 445 | Colborn/the Prize | prize | in-corpus (on-chain Name field) | collect | | 446 | SuperRareRoses | flower | weak alt to essence 'rose' | rose | CAVEATS: 4 of 11 words (champion/potato/mask/flower) rest on the agent's outside-the-corpus knowledge of real artist bios, not a mechanical deck rule — verify before trusting a hit. #447's word is literally its own Essence, which undercuts the "non-essence register" premise for that one slot. Net: 6 of 11 words are solid non-essence, in-corpus reads; 5 are weaker. Still genuinely the most information-bearing Tier-A set (9-10 distinct words vs. Silver-11's 1-4) and explicitly sidesteps the essence-register death. Promise: **high, with caveats**. ### A3. `S2-W4` (primary) + `S2-W3` (fallback) — Silver-SubType-11 Selection: all 11 pieces with `SubType == "Silver"` — `4,9,25,36,41,55,64,81,145,149,453` (verified exact; not a ratified glitch class, a plain trait selection). Mint order == GAN-id order (verified: mint 69,74,90,102,107,121,131,150,227,231,592). Rule: take Type word if BIP-39, else fall back to SubType word 'silver' (always qualifies). Recomputed Type values and BIP-39 membership directly: 4=Smoke(BIP)->smoke, 9=Glitch(not BIP)->silver, 25=Glitch->silver, 36=Smoke->smoke, 41=Glitch->silver, 55=Smoke->smoke, 64=Skull(BIP)->skull, 81=Ghost(BIP)->ghost, 145=Cyborg(not BIP)->silver, 149=Bot(not BIP)->silver, 453=Smoke->smoke. **Word set (mint order):** `smoke, silver, silver, smoke, silver, smoke, skull, ghost, silver, silver, smoke` Only 4 distinct words (smoke x4, silver x5, skull x1, ghost x1). Promise: **low** — worth one cheap pass since untested_methods_full.json itself flags this exact set as "silver-11 repeat unswept," but don't expect a hit. ### A4. `S2-W3` — pure Silver-SubType-11 Same 11 members as A3. Every SubType is literally 'Silver' -> the word is **`silver` x 11**, fully degenerate. BIP-39 permits word repeats, so it's not disqualified, only implausible. Promise: **lowest of the Tier-A set** — run it last, for completeness only. --- ## TIER B — needs a build (selector or partial word-register verified, not yet 11 words) ### B1. `EC40` — TWIN(13) SubType/Type transposition classifier (cleanest Tier-B lead) Recomputed directly against the deck: 11 of 13 TWIN members correctly self-declare their class term in SubType (`Twin`/`Twins`); exactly 2 invert it — Type carries `Twins` instead while SubType holds something else (#157 SubType=Binary/Type=Twins; #298 SubType=Slender/ Type=Twins). Confirmed exact: TRUE/KEEP = `47,59,73,121,124,132,139,323,373,379,395` (11), DROP = `157,298` (2). This is genuinely new (unflagged anywhere in the ledger) and gives a clean, mechanical, non-essence 13->11 cut. It is a **selector only** — needs a word register (owned by the S2-W* family) to become sweepable. Highest-confidence Tier-B item. ### B2. `S2-W17`+`OC17` — ENTANGLED filename-date numeric decode (10 of 15 resolved) Rule: ArweaveFilename trailing numeric token N (frame-position-in-season-transition, or `repriseNN`) -> bip39[N]. Recomputed directly from the deck's ArweaveFilename field — all values match exactly. Solstice subset: 217(N=8)->absurd, 222(N=1)->ability, 226(N=4)->above, 244(N=13)->accuse, 260(N=20)->action, 263(N=30)->adult, 281(N=62)->amazing. Review-reprise subset: 450(N=0)->abandon, 465(N=8)->absurd (repeat), 473(N=9)->abuse. 10/15 resolved (9 distinct + 1 repeat). Gap: `411,436` carry NO ArweaveFilename at all; `481,489,499` carry non-numeric tokens (`test`,`skullall`,`aa`) where a number should be — these 5 resist the mechanical rule and are the concrete next problem. Genuinely untested register on a ratified class's own authored vocabulary. Promise: **high** once the 5-piece gap is closed. Curiosity flagged by the source agent: 260's word "action" is the same word EC45 (below) flags as the sole essence-vocabulary hole — logged as a coincidence worth the collector's eye, not yet load-bearing. ### B3. `S2-W16` — filename-descriptor register on HODL/Named Fibonacci-eleven (6 of 11) Recomputed directly against ArweaveFilename — all confirmed exact: #1->ancient, #5->whale, #13->spirit, #34->heart, #55->warrior, #89->smooth (its own `_smooth` suffix, since 'shark' isn't BIP-39). Gaps (no BIP-39 descriptor found by token-scan): #0 (no filename — off-chain SuperRare anchor), #2 (charred), #3 (scholar), #8 (blob), #21 (beanie) — none of those 4 are BIP-39 words. Genuine thematic coherence on 6/11 ("ancient/warrior/spirit/heart/whale" reads as a real vocabulary family) but doesn't close without either visual inspection of the 5 gap pieces for a hidden second word, or a documented direct-synonym rule (must be a literal dictionary hit, not a cipher, to avoid reopening the banned word-transform class). Promise: **medium** — most coherent incomplete lead in the batch. ### B4. `OC17` — Day/Night calendar bookend on the ratified pairs+anchor set The one pair in D2's ratified pairs set with literal calendar semantics: #209 "Solar skullGAN" OnChain_Date=`Day` (verified), #300 "Lunar skullGAN" OnChain_Date=`Night` (verified) — the solstice drop's own first/last pieces. Rule: fix #209's word to position 1, #300's word to position 11, let the other 8 pairs' (+anchor's) word register fill the middle. Ordering-only — needs the pair-relation word register the pairs team is still building. Not previously ledgered as an ordering use of Day/Night (only the pair confirmation is recorded). Promise: **high** as a composable piece, not standalone. ### B5. `OC17` — ENTANGLED solstice-subset calendar order (6 of 15, real reordering effect) 6 of ENTANGLED's 15 members carry a real on-chain calendar date (all in the solstice sub-drop; verified: 217 Jun 11, 244 Aug 9, 222 Sep 15, 260 Oct 29, 226 Nov 5, 281 Dec 18). Sorting by day-of-year swaps two adjacent pairs vs. mint order — `217,244,222,260,226,281` instead of `217,222,226,244,260,281`. This is the cleanest demonstration in the batch that OC17 is a real, working, non-trivial mechanism on this data. Coverage is drop-local to solstice only (0 of 144 ghost-drop and 0 of 68 review-drop ENTANGLED members carry any OnChain_Date). Needs a non-essence word register to turn the order into words, and doesn't by itself reach all 15/11. Promise: **medium**. ### B6. `EC40` — ENTANGLED reprise-split, seed-pair-literal (9/6, doesn't close) Literal reading of the #363(TRUE)/#473(FALSE) seed pair: #473's Drop='Act 1 - Review' makes it a REPRISE of the solstice/ghost original; predicate = TRUE (native drop) vs FALSE (Review reprise). Applied to ENTANGLED(15) by Drop field: TRUE(solstice+ghost)=9 members, FALSE (review)=6 members. Neither is 11. #512 sits in the Review range so would read FALSE under this predicate — flagged tension with any #512-anchor idea, not silently patched. Promise: **medium** — grounded in the seed pair's own fields (unlike B1, which is an independently- derived analogous rule), but doesn't cut cleanly. ### B7. `S2-W21` — absence-word register (7 words from 5 absence-events, short of 11) Each authored absence contributes a word from its OWN drop-channel number/label (not Essence): #134 (missing gif, robo binary-position 5) -> absent; #446 (M_ArweaveCode field, different from Essence, literally 'roses') -> rose (note: happens to equal #446's own Essence too — flagged); #390 (double-mint slot 519) -> domain; identity-swap pair {14,75} -> achieve, annual; never-forged pair {49,20} -> alert, action. 7 words total, needs 4 more absence-events or a stack partner to reach 11. Two cross-register convergences flagged for the collector (both "action" and "domain" also surface independently in EC45/B2) — logged as curiosities given a 2048-word space, not yet signal. Promise: **medium**. --- ## TIER C — needs the collector's eye, or a confirmed negative worth keeping on record - **`EC47` invader-signature-without-the-name** — exact 11-member selection (`314,320,349,377,379,398,401,412,431,473,475`, contingent on excluding #340) on a ratified visual-tag corpus, genuinely novel. BUT the word half is unresolved AND the literalized homage-word itself, `invader`, is **confirmed not a valid BIP-39 word** (checked directly) — so even a clean census can't close this via the obvious word. Needs the collector's eye on whether the shirt/background icon actually varies piece-to-piece (current MichaelNotes read identically for all 11) before any word-derivation build is worth doing. - **`S2-W23` near-miss-homage register** — only 2 confirmed members (#357, #423; census "pending the collector's eye" per DECISIONS D4b/D6). Word candidates already logged and BIP-39-valid: #357 = yellow/blue/half/almost/shy/alien/spike; #423 = recycle/blue/green/ echo/mirror. Seed for a future census, not a ready recipe. - **`EC45` negative-space word scan** — recomputed and CONFIRMED exactly as claimed: across all 513 distinct Essence values, index 20 ("action") is the only unused word in bip39[0:225], and "seven" is the only unused word in the closed number-word sub-vocabulary. Only 2 words; tried extending the same "closed band, one hole" logic to family-relation and color word bands and both failed (multiple holes) — does not obviously generalize. Best used as a 2-word high-confidence addition stacked onto another class (e.g. B2 or B7). - **`S2-W9` painted-number index** — only the 2 D15/D16-canonical, collector-pinned values are trustworthy: #165's clearest reading (D16: "81" at f12/14/15, not the fuller "8941") -> bip[81]='any' (0-idx) or bip[80]='anxiety' (1-idx); #207 (D15: "4," not "#4") -> bip[4]= 'above' (0-idx) or bip[3]='about' (1-idx) — recomputed and confirmed. #179's D19 "maybe a 10.00" and #379's "AI" don't parse cleanly as radix pointers. 4-member painted-mark class ceiling — explicitly a stack contributor per D10 doctrine, never a standalone 11. - **Stretch tribute+near-miss+#512 combination** — reaches exactly 11 only by force-combining three separately-ratified classes; #282 and #423 would both plausibly read "recycle" (a legal but suspicious duplicate); #512's inclusion is contested (Guardrail G11). Weakest recipe in the batch, logged for completeness only. - **`EC40` TRUE/FALSE as marker-classifier (GLITCH_count proxy variant)** — tested predicate (FALSE-flavored = GLITCH_count>0) on ENTANGLED(15) gives 7/8, not 11. Needs a better- evidenced TRUE/FALSE predicate (drop/GLITCH_count/resave-session diff on #363 vs #473 itself was proposed as the next concrete step, not yet run). - **`EC47` inverse (invader-name-without-visual-confirmation)** — only 3 members (`328,342,510`), far short of 11, no known expansion path without a fresh full-collection visual census. Logged for ledger hygiene only. - **`OC17` whole solstice-drop calendar order (92 pieces)** — clean NEGATIVE: calendar order does not recover a more meaningful essence sequence than piece-ID order already gives (the piece-ID order is already the alphabetical BIP-39 dictionary run). One new structural fact surfaced (the date assignment "wraps" the year 6 separate times across the drop) — logged for the record, not a sweepable recipe. --- ## DEAD / CLOSED — cross-checked against the DEAD list or confirmed negative this pass - **`S2-W17` ENTANGLED month-word register** — only 3 of 15 members land on a valid BIP-39 month word (august/october/december); 8 of 15 have no date at all. Clean negative, distinct from the already-solved 4-piece cryptoArtGAN equinox/solstice set — don't re-try. - **`S2-W16` season-word variant on ENTANGLED** — only 3 of 15 give an unambiguous single season word (4 are transitional/tied, 8 have no filename). Same dead end as above. - **`S2-W11` drop-channel-medium register on backwards-Fib-eleven** — **overlaps-dead**: 6 of 11 members' "Leet-corrupted title" flag is just the piece's own binary-position suffix rendered 1/0-for-i/o, i.e. the already-ratified/DEAD Prime-Position binary-position register wearing a different label, not a new word source. The other 5 members carry no flag at all. - **`S2-W23`+`EC47` literalized homage-word "invader"** — confirmed `invader` is not a valid BIP-39 word; this register cannot supply a checksum-legal word under the obvious reading. See Tier-C `EC47` entry for the still-open selection question. - **`EC40` chain-vs-sheet corruption-flag classifier** on TWIN/ENTANGLED — flagged-contradiction counts (3 on TWIN, 1 on ENTANGLED) don't isolate an eleven either way; 2 of the 3 TWIN flags are trivial singular/plural mismatches, not real corruption. Exploratory only, not a working recipe. - **`S2-W3`/`S2-W4` literal trait register on TWIN(13)/ENTANGLED(15)/HiddenType-misfiles(4)** — heavily degenerate everywhere tried: 9/13 TWIN hits collapse to the single word 'twin', ENTANGLED misses 12 of 15 outright, HiddenType-misfiles is 'skull' x4 on a 4-member set (far short of 11 regardless). Carries almost no differentiating information; not worth a slot. - **`EC46` GAN/lore-vocabulary words on TWIN(13)/ENTANGLED(15)** — no principled 13->11 or 15->11 narrowing rule currently exists for either class (both already flagged as open in SOLUTION_SPACE.md), and the only words concretely findable via existing ledger content (echo, "space invader") mostly restate the already-dead essence register rather than adding new lore. Do not prioritize without a fresh narrowing rule first. - **`S2-W4` Type-word on HiddenType-misfiles(4)** — all 4 hit 'skull' uniformly (correct, but the selection itself is only 4 members, a building block at best, no combination proposed). --- ## Executive summary of ranking logic (for the sweep queue) 1. Run **A1** (both index variants) first — cheapest, most mechanically novel, only caution is the unresolved DEAD-list overlap question (worth a 5-minute registry/log check first). 2. Run **A2** as a parallel low-cost pass, but treat any hit with suspicion until the 4 external-knowledge words and the #447 essence-collision are independently re-verified. 3. Run **A3**/**A4** last — cheap, essentially sanity passes, low expected value. 4. Fast-track **B1** (TWIN transposition classifier) to the S2-W* word-register team — it's the cleanest unclaimed selector in the whole batch and just needs a word rule. 5. Fast-track **B2** (ENTANGLED filename-date decode) for a 5-piece gap-closing build — highest remaining-work-to-payoff ratio in Tier B.