133 files, 87 MB. Generated 2026-10-24. Machine-readable list at manifest.json. The briefing that explains these files is at /brief/; the live site is at bitganspuzzlebook.com.
Start here. These are his files and pages exactly as published. Everything else on this site is derived from them.
| file | MB | what it is |
|---|---|---|
| READ-ME-FIRST.md | 0.0 | Which files are ground truth, which are our work, what is known to be stale, and the one rule: nothing is closed. |
| all_onchain.json | 0.1 | On-chain trait pull, flat. |
| ar_hash_243.json | 0.0 | MD5 of every loose GIF on the drive, downloaded and hashed. |
| ar_zips.json | 0.1 | Contents of every drop archive on the drive: each GIF path, size, MD5 and the local piece it matches. |
| ardrive_entities.json | 0.1 | Every file/folder entity on Pindar's ArDrive (Arweave GraphQL, 13 Sep 2026). |
| ardrive_meta.json | 0.2 | ArFS metadata for each entity: name, size, dataTxId, lastModifiedDate. |
| arweave_bitgan_properties.csv | 0.0 | Pindar's original Arweave properties sheet, unchanged. |
| arweave_file_txids.json | 0.0 | Arweave transaction id per file. |
| bip39.txt | 0.0 | The 2,048-word BIP-39 English list. |
| burn_all.json | 0.0 | Burn transactions. |
| chain_descriptions.json | 0.2 | The FULL on-chain description for all 512 pieces, from the OpenSea tokenURI endpoint. Blockscout's cached copy is incomplete (111 of 512 return no traits) and truncates to 120 characters. |
| chain_raw_by_mintindex.json | 0.3 | On-chain traits keyed by mint index. |
| journal/journal_hidden-lore-revealed.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_le-crucible.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page002.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page003.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page004.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page005.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page006.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page007.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page013.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page014.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page015.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page016.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/journal_page017.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/news_cryptoartgans.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/news_hidden-types.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/news_pandimensional-skullgan-ch2.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/news_pattern-revealed.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| journal/news_the-final-bitgan-drop.txt | 0.0 | Text of a bitgans.com journal or news page, saved 13 Sep 2026. |
| onchain_meta.json | 0.1 | On-chain metadata per token. |
| page015_jan1/page015_capture_2022-01-01.html | 2.3 | The 1 January 2022 Arweave capture of journal page 015, reachable only through piece #272's on-chain description. Contains an inline 300-row properties table. |
| page015_jan1/page015_jan1_table.csv | 0.0 | That table extracted: pieces 1-300, with the four columns that exist nowhere else (Decimal, Hidden Type, Strokes, bip-0039). |
| transfers.json | 0.5 | Every on-chain transfer of all 512 tokens (Blockscout), with timestamps. |
Reproducible by anyone from the tier above: hashes, frame counts, save times, transfers, number classes. primary_facts.py is the code.
| file | MB | what it is |
|---|---|---|
| ARWEAVE-LAYER.md | 0.0 | The rule behind each drop's filenames, dates and seed-list indices on Arweave, and every place a piece breaks it. |
| DROP-NUMBER-RULES.md | 0.0 | The number rules behind each drop's looks (Pindar's number classes, rhythms and binary patterns), and the ten pieces that break them. |
| arweave_filenames.json | 0.0 | Each piece's filename as it appears in Pindar's Arweave archives (all 512 byte-identical to our copies). |
| burnt_gans_set.json | 0.0 | The burnt pieces. |
| chain_complete.json | 0.2 | Chain coverage record. |
| fib_return_frames.json | 0.0 | For the eleven backwards-Fibonacci pieces and #291: frame count, the frame positions where the returning frame appears, and the gaps between visits (5, 3, 2, 1, 1). Measured 19 Sept 2026 by near-match on downsized frames. |
| filename_descriptors.json | 0.0 | Arweave filename words not in metadata. |
| first_transfers.json | 0.0 | Each piece's first transfer out of Pindar's wallet (the effective mint). |
| gif_facts.json | 0.3 | Per-file measurements of all 512 GIFs: frame count, per-frame delays, md5, size. |
| glitch_instances.json | 0.0 | Every sheet-versus-chain difference listed as its own glitch: 34 glitches on 29 pieces, each with the field, both values, the kind of change, which side is odd, and the three-witness verdict. Built 19 Sept 2026 from primary_facts.py. |
| global_numclass.json | 0.0 | Number class of every piece by GAN and mint number. |
| page015_jan1/bip39_index.json | 0.0 | BIP-39 index per piece. 1-300 are Pindar's own bip-0039 column; 301-512 derived from the published Essence. |
| primary_facts.py | 0.0 | Recomputes every chain/file/number fact from the primary sources (the module the site build uses). |
| review_invariant_match.json | 0.0 | For every Act 1 Review piece, the five closest earlier pieces under eight colour-blind measures (outlines, brightness, brightness with inversion allowed, colour intensity, each also mirrored), with scores from 0 to 1. A true copy would score about 0.95; the highest anywhere is 0.70. |
| review_reprise_map.json | 0.0 | Every Act 1 Review piece that shares its exact title with an earlier piece, and the earlier pieces it matches. Thirteen match exactly one. Built 19 Sept 2026. |
| same_name_odd_ones_out.json | 0.0 | For every name shared by three or more pieces: the piece that is the single odd one out on a trait (drop, eyes, dimensions, Type, SubType, species, frame glitch), and what the others have. |
| same_title_groups.json | 0.0 | Every name shared by two or more pieces, with the piece numbers. 37 names are shared by exactly two pieces. |
| save_sessions.json | 0.1 | Every file's Photoshop save time and build, clustered into 39 sessions (3-hour gap rule). |
| so_analogues.json | 0.0 | For each of the eleven so-words after solution: its closest pieces in drops one to five by data (same name, same traits, binary counter) and by four machine-vision measures, with scores. |
| solution_key.json | 0.0 | The twelve-word solution block read as a reshuffle key: the wordlist rank of each so-word taken in piece order (1, 2, 10, 6, 3, 4, 5, 8, 7, 9, 11, 12), and the same with the heading removed. |
| tags_outside_themes.json | 0.0 | Tags that sit in no broad kind of glitch, mostly descriptions. |
| theme_sides.json | 0.0 | Each broad kind of glitch split by side (from the data, by eye) with its piece count and the first piece that shows it. |
| token_index.json | 0.0 | Mint index per piece (for token-id reconstruction). |
| xmp_ids.json | 0.2 | Photoshop XMP identifiers per file with the save timestamp decoded from the UUID. |
| zip_mtimes.json | 0.1 | Each archived GIF's disk modification time from the Arweave zips against its Photoshop save time. |
Interpretation: the anomaly ledger, the taxonomy, the collector's eye. Well-evidenced, but it is a reading, and it has been corrected before.
| file | MB | what it is |
|---|---|---|
| NAMING-CONVENTIONS.md | 0.0 | How Pindar builds a title from the GANg, Type and SubType fields, which drops follow the formula, and every title that breaks it. |
| anomaly_ledger.json | 1.4 | ~2,350 per-piece findings (type, desc, source). Append-only. |
| confirmed_glitch_registry.md | 0.0 | Confirmed glitches with evidence. |
| contract_forensics.md | 0.0 | Contract and token-id forensics. |
| creature_census.json | 0.0 | Visible-creature census (partly collector-corrected). |
| creature_census_report.md | 0.0 | Creature census report. |
| deck_512.json | 1.4 | Master dataset as of 13 Sep 2026: one wide record per piece with exact-case chain names, collector notes and derived columns. Does not carry the tags added 16–17 Sep (the January-table findings); the live record is /api/pieces. |
| forge_mapping.md | 0.0 | Forge pairs and burns. |
| gan_lore_expanded.md | 0.1 | Lore notes per drop. |
| glitch_dictionary.json | 0.1 | Every anomaly type in the ledger with a definition and a confidence. Extended 17 Sep 2026 to cover the types added by the audit and the January table. |
| michael_categories.json | 0.0 | His notes grouped by category. |
| michael_visual_notes.json | 0.0 | The collector's hand notes on all 512 pieces. |
| our_spreadsheet.csv | 0.3 | Our working spreadsheet, flat CSV, 68 columns under two header rows (group, then column): traits, mint, burnt, filenames, sheet-versus-chain verdicts, first transfers, save sessions, frame counts, hidden-type meaning, number classes. Rebuilt on every site update. The tag-by-tag version is /api/export.csv. |
| pairs.json | 0.0 | The eleven deliberate pairs with reasons. |
| puzzle_pages.md | 0.0 | Pindar's puzzle pages, transcribed and analysed. |
| sheet.json | 0.8 | Our working spreadsheet as JSON keyed by piece number, snapshot of 13 Sep 2026. The live record is /api/pieces and /api/export.csv. |
| signature_dictionary.md | 0.0 | Every name's visual rule and its violators. |
| taxonomy.json | 0.0 | How anomaly types fold into site concepts. |
| verify_d50_report.md | 0.0 | Second-pass verification of the non-visual tags against primary data (13 Sep 2026). |
| visual_audit_2026-09-24.json | 0.0 | The collector's frame-by-frame visual audit of the skull and robo drops, 24 September 2026: 117 notes on 192 pieces. |
| visual_audit_2026-09-25.json | 0.0 | The collector's frame-by-frame visual audit of the skull, robo and spooky drops, 25 September 2026: 133 notes on 208 pieces. Supersedes the 23 and 24 September files. |
| visual_audit_2026-09-27c.json | 0.1 | The collector's complete frame-by-frame visual audit of all six drops, 27 September 2026: notes on pieces 1 to 512. Supersedes the earlier audit files. |
| visual_audit_skull_2026-09-23.json | 0.0 | The collector's frame-by-frame visual audit of the skull drop, 23 September 2026: 71 notes on 128 pieces. |
Recipes, sweeps, brainstorms, hand-offs and rulings, in date order. Nothing here is closed: a negative sweep says one recipe missed, and several were run on data we have since corrected.
| file | MB | what it is |
|---|---|---|
| AUDIT-BRIEF-2026-09-14.md | 0.0 | The brief handed to an independent auditor on 14 Sep 2026: 39 claims about the rebuild, with sources. Its claim 3 (the noungans test zip) and its note that no earlier sheet revision exists on the drive were both overturned — see AUDIT-REPORT and FINDINGS-2026-09-16. |
| AUDIT-REPORT-2026-09-14.md | 0.0 | The independent auditor's report on the 39 claims: 31 stand, 8 do not. |
| BRAINSTORM_32.md | 0.0 | The 24-idea "32" board with results. |
| DECISIONS.md | 0.4 | The rulings log, D1–D70, in date order. The later entries correct earlier ones; read a ruling with its date. |
| DEEP-REVIEW-LIST.md | 0.0 | The standing list of every 9-to-12 piece set to review in depth with the collector: what it is, which side it is on (data or visual), whether it has been run, and whether its review page exists. |
| FINDINGS-2026-08-31.md | 0.0 | Consolidated findings of the lore-catalog run-down, 31 Aug 2026. Its 'dead' verdicts are per-recipe notes, not verdicts on the glitches (D70). |
| FINDINGS-2026-09-16.md | 0.0 | The January properties table: how #272's odd on-chain link leads to Pindar's own bip-0039 index column, and what its four extra columns settle. |
| FIRST-PRINCIPLES-BRIEF-2026-09-27.md | 0.0 | Self-contained brief for a fresh run at the puzzle from first principles: the clue, the collector's tests and rules, the logic of every drop with its breaks, cross-drop facts, a swarm plan, open hypotheses, and the data map. |
| HANDOFF-2026-08-26.md | 0.0 | Earlier handoff (harness and hard rules). |
| HANDOFF-2026-08-28.md | 0.0 | Working hand-off of 28 Aug 2026: goal, rules, state, materials. Its 'closed / dead' section is history — under D70 nothing is closed, and several of those runs used data corrected since. |
| HYPOTHESES-DRAFT-2026-09-16.md | 0.0 | The January-table hypotheses in the project formula (selection / word / ordering), and the 16 Sep 2026 run results: 20 orderings, 2,560 seeds, plant seen, no hit. |
| LORE-CATALOG-TRIAGE-2026-08-31.md | 0.0 | Triage of the 46-item GAN-culture catalog. |
| METHODS_RECIPES.md | 0.0 | Untested-methods swarm, ranked recipes. |
| REGISTRY_POLICY.md | 0.0 | How the no-hit registry is used (26 Aug 2026): a registry hit is a de-prioritisation hint, never a hard skip. This is the rule D70 restates for the whole site. |
| SOLUTION_SPACE.md | 0.0 | Glitch catalog, the 47 word rules, the 33 orderings, guardrails (Aug 2026). Its 'dead' marks are per-triple notes, as the file itself says; read with D70. |
| TEACHING_RECIPES.md | 0.0 | "Glitches teach the method" swarm synthesis. |
| chain_mismatches.json | 0.0 | SUPERSEDED snapshot (13 Sep 2026) of raw sheet-vs-chain differences, including title-counter noise. The ruled analysis is the 26-piece 'Sheet and chain disagree' tag family on the site and verify_d50_report.md; kept for history. |
| coverage.json | 0.0 | The hypothesis space as a grid: every (selection, extraction, ordering) triple the 44 sweeps touched, with its result, plus the selections, extractions and orderings never swept at all. Built for LLMs; the Hypotheses tab renders the same grid for people. |
| curated_elevens.md | 0.0 | Curated candidate elevens (Aug–Sep 2026). Its 'seven damaged essence slots' (#21, #57, #191, #259, #268, #363, #473) anticipated the January-table finding; the four robots that hold block words complete that set to eleven (SEL-SCHEME-EXCEPTIONS). |
| glitch_pool.json | 0.0 | Small authored classes for mix-and-match recipes. |
| instruction_reading.md | 0.0 | Glitch-as-instruction analysis. |
| instruction_schemes.json | 0.0 | Consistent-operand schemes for the verbs. |
| interpretation_matrix.json | 0.0 | 169 readings of the corrupt eleven's error verbs as operations. |
| journal_reread_2026-08-26.md | 0.0 | Re-read of Pindar's journal. |
| methods.json | 0.1 | SNAPSHOT of 28 Aug 2026: 27 selections, 47 word rules, 33 orderings with the statuses they had then (twenty free-text values). The live, current versions are /api/principles and coverage.json; selection statuses were normalised to four words on 17 Sep (D69). Kept for history. |
| mirror_report.md | 0.0 | Review-vs-earlier mirror search report. |
| multistep_chains.md | 0.0 | Multi-step recipe ideas. |
| pb_content.json | 0.1 | Plain-English prose and Pindar rationale per method, written 28 Aug 2026. Its colour-word entry predates the 17 Sep widening (rose #446 and bronze #493 are colour words too). |
| runs_2026-09-16/hypotheses.py | 0.0 | The exact word lists and orderings swept on 16 Sep 2026. |
| runs_2026-09-16/run_backfill_2026-09-17.json | 0.0 | The hand back-fill that gave each of the 44 sweep runs its selection / extraction / ordering ids, read from the run's own label and notes. |
| swarm_2026-09-21/CONTEXT.md | 0.0 | The briefing given to the five idea agents on 21 September 2026: the clue, the rules, the collector's rulings, and what had already been tested. |
| swarm_2026-09-21/lens1_same_names.md | 0.0 | Idea agent 1: what the same-name groups could mean. Six ranked ideas. |
| swarm_2026-09-21/lens2_lying_labels_instructions.md | 0.0 | Idea agent 2: labels that lie about the picture, and instruction words as a program. Six ranked ideas. |
| swarm_2026-09-21/lens3_review_answer_key.md | 0.0 | Idea agent 3: Act 1 Review as the answer key. Six ranked ideas. |
| swarm_2026-09-21/lens4_puzzle_design.md | 0.0 | Idea agent 4: how the artist would have had to build the puzzle, and cipher history. Six ranked ideas. |
| swarm_2026-09-21/lens5_order.md | 0.0 | Idea agent 5: what could fix the order of the words. Six ranked ideas and the dead ends it computed. |
| tested_sets_no_hit.txt | 10.4 | Every eleven-word set a sweep has checked against the wallet with no hit (146,486 sorted sets), each only in the orderings its run stated, × 128 checksum-legal twelfth words × 33 paths. A coverage record, not a verdict: some were swept on data we have since corrected, and the policy (D70) is that nothing is closed. |
| tested_sets_no_hit_2026-09-21.txt | 17.4 | Every phrase tested on and after 21 September 2026 that found nothing: the eleven words in order, then what the test was, in plain words. New tests are added here. |
| tested_sets_no_hit_2026-09-28.txt | 8.3 | Eleven-word sets swept with no hit that the other tested-set files did not hold (117,146 sorted sets): those swept by the second swarm attempt of 27 and 28 September 2026 (cards 282 to 406, rounds 1 to 8), and the rest from the older registry and earlier plant-validated runs recorded in the swarm's coverage file. A coverage record, not a verdict. |
| tested_sets_no_hit_2026-09.txt | 20.1 | Every phrase tested on 19 and 20 September 2026 that found nothing: the eleven words in order, then what the test was, in plain words. About 80,000 phrases. Split from tested_sets_no_hit.txt because one file may not exceed 25 MB. |
| word_sources.md | 0.0 | Every word source considered. |
Large HTML pages that show the evidence for one topic. Read them as of their date.
| file | MB | what it is |
|---|---|---|
| brainstorm32_board.html | 0.0 | The 32 decoder board. |
| corrupt11_rows.html | 9.0 | Explainer: the corrupt eleven row by row + the eleven pairs (two tabs, embedded animations). |
| creature_review.html | 6.2 | The creature review the collector adjudicated. |
| glitch_verbs.html | 0.1 | Explainer: every reading of the glitch verbs. |
| glitch_verbs_by_logic.html | 0.1 | Explainer: every verb reading grouped by operand logic. |
| hiddentype_primer.html | 4.7 | Explainer: the Hidden-Type misfiles, with embedded animations (Sep 2026). Written before the January table showed that #32 and #119 are chain-side errors Pindar had fixed in his own sheet, while #161, #187 and #193 are carried deliberately. |
The sweep harness and the fact recomputer.
| file | MB | what it is |
|---|---|---|
| build_tags.py | 0.0 | Builds every tag assignment and its evidence line from tag_schema.py and primary_facts.py. |
| harness_qbuild.py | 0.0 | Builds a plant-validated sweep queue. |
| harness_qrun2.py | 0.0 | Runs a queue; exits 99 on a real hit. |
| harness_sweeplib.py | 0.0 | Sweep harness: BIP-39 checksum, 33 derivation paths, plant. |