# The Act 1 Puzzle page, mined completely — and every bitgans.com page published on or before 5 February 2022

Worker report. Scope: the bitgans.com site archive in `site/` (109 pages, `site/raw/*.html`
and `site/txt/*.txt`, plus `site/meta.json` and `site/sitemap.xml`). Method: raw-HTML
inspection, BIP-39 word extraction against the 2048-word English list (`from fast3 import WL,
WIDX`), and a control pass over the two puzzles Pindar has already solved in public.

---

## 1. The headline result

The Act 1 Puzzle page is `https://bitgans.com/news/cryptoartgans`, archived locally as
`site/raw/news__cryptoartgans.html` and `site/txt/news__cryptoartgans.txt`. Its embedded
metadata gives `datePublished = 2022-02-05T06:12:55-0600` and `dateModified =
2022-02-05T06:47:15-0600`, so it was published and then touched once, 35 minutes later, and
has never been edited since.

**The roboGAN method does not transfer to this page.** The body of the Act 1 Puzzle page
contains 45 BIP-39 word tokens, of which 34 are distinct. It does not contain 11 BIP-39
words, and no subset of 11 is picked out by any typographic device, because the page contains
no typographic devices at all beyond a single `<strong>` on the clue sentence itself. There is
no red text, no coloured span, no hidden element, no HTML comment, no `alt` text, no `title`
attribute, no anchor name, no `data-` payload, no inline `<script>` or `<style>` carrying
content, and no image whose filename is a word. I checked all of these explicitly and the page
is clean on every one.

**Every candidate 11-word sequence I could extract from this page by a stateable rule has been
tested against the target wallet and all of them fail.** I enumerated the natural extraction
rules (first 11, last 11, first 11 distinct, last 11 distinct, first 11 once-only words,
paragraph unions that happen to total 11), computed for each the 128 checksum-valid twelfth
words, and derived addresses on the five common Ethereum paths
(`m/44'/60'/0'/0/0`, `m/44'/60'/0'/0`, `m/0/0`, `m/44'/60'/0'/0/1`, `m/44'/60'/1'/0/0`).
That is 6,400 derivations for the forward orders and a further 13,440 for reversed,
alphabetical and BIP-39-index orderings of the seven page-derived sets. Every one missed
`0x18f87ec9c527aba1db44f715456bf28b0dae478d`. The sets are still recorded in
`theories/puzzle_pages_sets.json` so nobody repeats the work, but they should be treated as
already eliminated on those paths, not as live leads.

**More importantly, the premise that motivated this task needs correcting.** The roboGAN
puzzle's answer was not in the prose of the clue page. It was in an attached image and a CSV
of *artwork properties*. Section 3 below sets this out from Pindar's own words. Once that is
established, all three solved puzzles have the same shape — the words come from the artworks'
metadata, and the ordering comes from mint order — and none of them has ever drawn a single
seed word from the text of a news post. That is the strongest thing this page-level
investigation has to say: the answer to the Act 1 Puzzle is almost certainly not on any page.

---

## 2. The Act 1 Puzzle page in full

### 2.1 The complete rendered text

Title: **cryptoArtGANs**. Dateline: **Feb 5**. Byline: **Pindar Van Arman**.

> The bitGANs have created themselves in the likeness of four more cryptoArtists. Each a big
> influence on their creation. - StellaBelle, Colborn, SuperRareRoses and Brinkman -
>
> The first three have already adopted their new owners. But the final cryptoArtGAN, has
> disappeared into a new mystery wallet at:
>
> https://opensea.io/0x18f87ec9c527aba1db44f715456bf28b0dae478d
>
> As far as cracking the wallet holding the Magic Internet moneyGAN. This one will be
> extremely hard, probably harder than all previous bitGAN puzzles.
>
> Only 1 clue will ever be given, even if this takes over a year to solve. So advanced warning
> that this is much harder than any of the previous puzzles, including the one that took a
> month to solve by the entire discord. Please budget your time and expectations appropriately
> because no clues beyond this first one will ever be given out.
>
> **The Act 1 Puzzle:**
> **The seed phrase is 11 glitches ending with a random word.**

That is the entire authored content. The page is 149 whitespace-delimited words and 728
letters long. It is built from four Squarespace text blocks and four image blocks.

### 2.2 What the raw HTML contains, block by block

The article region of `news__cryptoartgans.html` is 40,187 characters inside a 389,417-character
file; everything else is Squarespace chrome. In document order the article holds:

1. A text block with the "four more cryptoArtists" paragraph. No inline markup at all.
2. An image block: `st.gif` (512×512), `alt=""`.
3. A text block with the "mystery wallet" paragraph, containing the raw OpenSea URL as plain
   text — it is **not** a hyperlink; there is no `<a>` around it.
4. An image block: `br.gif` (512×512), `alt=""`.
5. An image block: `ro.gif` (512×512), `alt=""`.
6. An image block: `co.gif` (512×512), `alt=""`.
7. A text block with the "extremely hard" and "Only 1 clue" paragraphs, joined by `<br><br>`.
8. A text block containing the clue, centred, in the `sqsrte-large` class, wrapped in
   `<strong>`: `The Act 1 Puzzle: <br>The seed phrase is 11 glitches ending with a random word.`

The only two links anywhere in the article are the author link `/?author=55e57fb9e4b029d4f2d0dfe1`,
which appears twice and is template boilerplate.

### 2.3 Everything the raw HTML hides or emphasises

Emphasis: exactly one element. `<strong>The Act 1 Puzzle: The seed phrase is 11 glitches
ending with a random word.</strong>`. Nothing else in the article is bold, italic, underlined,
marked, a heading, superscript, subscript or code. Search for `<span style="color:...">`
returns nothing; a search for any `color:` declaration in an article-level inline style returns
nothing.

Hidden content: none. There is no `display:none`, no `visibility:hidden`, no `opacity:0`, no
zero-size element and no text coloured to match the background anywhere in the article.

HTML comments: the file contains 18 non-empty comments and all 18 are Squarespace template
labels — "This is Squarespace.", "Background", "Social", "Title and nav wrapper", "Title",
"Nav", "Actions", "Burger", "(Mobile) Menu Navigation", "Menu Navigation", "End of Squarespace
Headers". None is authored.

`data-` attributes: the article uses thirty distinct `data-` keys and every one is a standard
Squarespace attribute (`data-block-type`, `data-image-dimensions`, `data-section-id` and so
on). None carries a message.

Embedded JSON: `Static.SQUARESPACE_CONTEXT` on this page exposes an `item` object with only
six keys — `commentState`, `fullUrl`, `id`, `publicCommentCount`, `recordType`, `title`. There
is no `tags` array, no `categories` array, no `excerpt` and no alternate `body`. The item id is
`61fe69c7e6479d42cb5de985`. Public comment count is 0.

Meta tags: `og:description` and `twitter:description` both simply repeat the first paragraph.
`og:title` is "cryptoArtGANs — bitGANs". `<meta name="description">` is empty.

Images: the four in-body GIFs are named for the four artists — `st.gif` for StellaBelle,
`br.gif` for Brinkman, `ro.gif` for SuperRareRoses, `co.gif` for Colborn. All four carry
`alt=""`. **The one small oddity on the whole page is that the images are not in the order the
text names the artists.** The prose says "StellaBelle, Colborn, SuperRareRoses and Brinkman";
the images run st, br, ro, co. That is a four-element permutation and it cannot by itself
produce eleven of anything, but if someone later needs an ordering hint from this page, this is
the only ordering discrepancy it contains.

The social-card image is different again and is not displayed in the body:
`gitm_3__1001.gif`, served from `static1.squarespace.com`, 334×334. "gitm" plausibly stands for
the Magic Internet moneyGAN named in the text, and `1001` looks like an identifier. This is the
only asset on the page that a reader never sees. It is worth someone pulling that GIF and
looking at its frames and metadata, given that the byteGAN puzzle was later solved by finding
seed words inscribed faintly across a GIF's twelve frames and duplicated in its Photoshop layer
metadata. I did not fetch it because it lives on `static1.squarespace.com`, outside the
`bitgans.com` fetch boundary I was given.

`site/sitemap.xml` lists the page with `lastmod 2022-02-05` and the four GIFs, each with the
Squarespace default title "News - cryptoArtGANs - Make it stand out" and the default caption
"Whatever it is, the way you tell your story online can make all the difference." Both strings
are template defaults that appear on every image in the sitemap, including on the two solved
puzzles' pages, so they carry no information.

### 2.4 The complete ordered list of BIP-39 words in the page

Reading the authored body in document order and taking every whole word that is an exact
member of the BIP-39 English list, case-folded, gives 45 tokens:

```
have, more, first, three, have, already, final, into, mystery, magic, this, one, will,
hard, all, only, will, this, over, year, solve, that, this, much, any, one, that, month,
solve, entire, please, budget, time, because, beyond, this, first, one, will, act, puzzle,
seed, phrase, random, word
```

Deduplicated in order of first appearance, 34 distinct words:

```
have, more, first, three, already, final, into, mystery, magic, this, one, will, hard, all,
only, over, year, solve, that, much, any, month, entire, please, budget, time, because,
beyond, act, puzzle, seed, phrase, random, word
```

If the title, dateline and byline are included the count rises to 47 tokens and 35 distinct,
the extra word being `van` from "Pindar Van Arman", which appears twice.

Broken down by paragraph, the counts are: paragraph 1 gives 2 words (`have`, `more`);
paragraph 2 gives 7 (`first three have already final into mystery`); paragraph 3 gives 6
(`magic this one will hard all`); paragraph 4 gives 24 tokens and 18 distinct; the heading
line "The Act 1 Puzzle:" gives 2 (`act`, `puzzle`); and the clue sentence itself gives 4
(`seed`, `phrase`, `random`, `word`).

**Applying the roboGAN rule literally to the clue — "the only BIP39 words contained in the
clue" — yields four words, not eleven: `seed phrase random word`.** Applying it to the clue
plus its heading yields six. Applying it to the whole page yields 34. There is no reading of
"the clue" on which this page contains exactly eleven BIP-39 words.

### 2.5 Structural counts, near-misses and acrostics

The page has six paragraphs whose initial letters spell `TTHAOTT`, and nine sentences whose
initial letters spell `TE-BTOSPT`. Neither is a word, an anagram of anything useful, or a
BIP-39 word. First letters of the four image filenames give `sbrc`; in text order they give
`scrb`. Nothing.

The near-misses on eleven are worth recording precisely, because they are the sort of thing a
later worker will rediscover and get excited about:

- Restricting to BIP-39 words of six or more letters gives exactly **ten** distinct words:
  `already, mystery, entire, please, budget, because, beyond, puzzle, phrase, random`. One
  short. No length threshold gives eleven: at five-plus letters the count is 16, at seven-plus
  it is 3.
- The **last eleven BIP-39 tokens** of the body are `beyond this first one will act puzzle
  seed phrase random word`, which ends on `random word` — literally the two words the clue says
  the phrase ends with. This is the single most seductive coincidence on the page and it is
  a coincidence: the clue sentence ends with the phrase "a random word", so of course the last
  two BIP-39 tokens before it are "random" and "word". It fails against the wallet.
- The union of the mystery-wallet paragraph and the clue sentence gives exactly eleven:
  `first three have already final into mystery seed phrase random word`. Also fails.
- Capitalised BIP-39 words on the page number six (`magic, this, only, please, act, puzzle`),
  not eleven.

### 2.6 Cross-checking the page's words against bitGAN essences

Since every bitGAN carries a BIP-39 word in its Essence property, I checked whether the page's
words are disproportionately drawn from the 512 essences actually used in Act 1 (source:
`artifact_items.json`, 512 items, 512 distinct essences). Twenty-two of the page's 34 distinct
words are essences, a rate of 0.63. That looks striking until you run the same test on every
other page in the archive: the median rate across 101 pages is 0.40, and `news/a-puzzling-collab`
scores 0.62, `journal/page006` 0.60 and `news/anotherpainting` 0.60. The Act 1 Puzzle page tops
the list but is inside the ordinary spread, and the elevated baseline everywhere is simply
because the essence list contains many short common English words that turn up in any prose.
**This is a negative result: the page's vocabulary is not secretly a list of essences.**

Narrowing further, only four of those twenty-two words are the essence of a bitGAN whose Type
or SubType is "Glitch" — `three` (g176 Slenderbot roboGAN), `already` (g58 Rainbow Glitch
skullGAN), `final` (g512 Quantum glitchGAN) and `that` (g158 G1itch1ng cryptoGAN). Four, not
eleven. For reference the collection contains 67 Glitch-typed items.

---

## 3. The control: what the two solved puzzles actually did

### 3.1 The roboGAN puzzle — `news/the-puzzle`, "The Puzzle (solved)", 28 September 2021

The full authored text of that page is short:

> Congrats to noohp who has solved the riddle in under 3 hours.
> **Solution is at end of this post**
>
> [roboGANproperties.csv]  ← an `<h1>` linking to a Google Drive file
>
> ## Solution
> The first step of the solution was to realize that the 12 bip39 words used for the seed
> phrase for the hidden wallet, were the only BIP39 words contained in the clue. (In red below)
>
> The second step was to use the minting order of the bitGANs to determine the order of these
> 12 words. Congrats to noohp for figuring this out within 3 hours.

The page carries three images, in this order: `clue_subtle_888.gif`, `csv_list.png`, and
`clue_solution.gif`. The social-card image is `clue_subtle_888.gif`.

This is the crux. **"The clue" is one of those images, and "(In red below)" points at another
image — not at the page's own prose.** The page's own prose contains 22 BIP-39 tokens and 17
distinct, nowhere near 12, and none of them is highlighted in the HTML: the only emphasis on
the entire page is `<strong>Solution is at end of this post</strong>` and three headings
(`roboGANproperties.csv`, `Solution`, and the congratulations line). There is no coloured span
in the HTML of this page either. The red highlighting that Pindar refers to exists only inside
`clue_solution.gif` or `csv_list.png` as pixels.

The `roboGANproperties.csv` link, and the corroborating account on `news/superrare-bitgans`
("noohp was able to crack the puzzle by discovering a seed phrase hidden in their properties"),
settle what "the clue" was: the roboGANs' own property table. The seed words were property
values, and mint order supplied the sequence.

I tested the obvious version of this against the local metadata as a sanity check. Among
bitGANs 129–192 (the 64 renegades, drop label `2021.9.29.1202`), only seven have titles
containing a BIP-39 word — `awkward`, `cat`, `slender` (twice), `dog`, `evil`, `olive`, `wolf`
— so the twelve did not come from the Name column alone. Establishing exactly which column
they came from needs the CSV or the two clue GIFs, and **neither the CSV nor
`clue_subtle_888.gif` nor `clue_solution.gif` is in the local archive**. Recovering them is the
single highest-value missing artefact for understanding his method, and I recommend someone
with the wider fetch permission pull all three.

### 3.2 The Solstice puzzle — `news/pattern-revealed`, "Pattern Revealed", 14 January 2022

Pindar's own account:

> Here is how bitGANs hid the seed phrase. Their nature was seasonal, so it was all in the date
> properties. If you looked at nothing but the date property, you could find 12 seeds words,
> some hidden and others in plain sight. The first 2 seed phrases were in plain site, "day" and
> "night." Then 4 were hidden in a date... And the next 6 were hidden in plain sight as the 6
> months that are also bip39 words... And finally, their order was according to mint order.

The answer, set in `<strong>` on that page, is:

```
day spring march april august summer october autumn december winter february night
```

Two things matter here as control. First, the twelve words came from **one metadata property
across 92 artworks**, and again the ordering was **mint order**. Second, this page is the one
place in the entire archive where a seed phrase appears in bold — twelve BIP-39 words inside a
single `<strong>` — which tells you what Pindar's bold actually means: **bold marks the
answer, on a solution page, after the fact.** On the Act 1 Puzzle page bold marks the clue, not
the answer.

### 3.3 The other two puzzles, for completeness

`news/podgallery-solution` (3 December 2021) describes a puzzle where the twelve words came
from the podGANs' **attributes**, with a twist: the attribute values were SLIP-39 words and had
to be re-indexed into BIP-39. The ordering came from the first-and-last square of each of the
six pods. Metadata again.

`news/the-bytegan-puzzle` (3 March 2023, well after the cutoff but methodologically important)
records that bitquence found the seed phrase in the **Photoshop layer metadata of an on-chain
GIF**, while noohp independently found the same twelve words "inscribed very lightly in the
background" of the GIF's twelve frames. Pindar adds the ruling that matters: "The cool thing
about seedphrase puzzles, is it doesn't matter how you find the seedphrase and there is no
judge."

### 3.4 What the glitch page shares with the solved pages, and what is unique to it

Shared: the announcement pattern (an artwork has absconded to a named wallet, the wallet
address is given as plain text, a clue follows), the one-sentence clue in bold, and the
absence of any HTML-level trickery. In that structural sense the Act 1 Puzzle page is a
completely ordinary bitgans.com puzzle announcement.

Unique to it, and there are four things:

1. It is the only puzzle post that says up front that **no further clue will ever be given**,
   and it says so twice in one paragraph.
2. It is the only puzzle whose clue names a **count of source artworks (11)** and explicitly
   marks the twelfth word as **random**, that is, as carrying no information. Every other
   puzzle produced all twelve words from the artworks.
3. It is the only puzzle in the archive that **never received a solution post**. A full-text
   search of all 109 pages for `0x18f87ec9`, `moneyGAN`, `11 glitch`, `Act 1 Puzzle` and
   `Magic Internet` returns `news/cryptoartgans` and nothing else. Every other puzzle has a
   matching "solved" page; `news/the-bytegan-puzzle` even indexes four of them, and this one is
   not on that list.
4. Its social-card image, `gitm_3__1001.gif`, is the only asset on the page that is never
   displayed to a reader.

---

## 4. Every page published on or before 5 February 2022

Seventy-four dated pages fall on or before the cutoff, plus twelve undated pages (site
furniture and unused Squarespace template stubs). Dates are taken from each page's own
`<meta itemprop="datePublished">`, which agrees with `site/meta.json` throughout; the
`bytegans__project-*` stubs and the top-level index pages carry no date in either source. The
"Title" column is the first line of the text extract, which for a few image-led pages is an
image caption rather than a headline. "Emphasis tags" counts `<strong>`, `<b>`, `<em>`, `<u>`
and `<mark>` in the article region.

| Date | Page | Title | BIP-39 tokens | distinct | emphasis tags |
|---|---|---|---|---|---|
| (undated) | `about` | The Artist (bitGAN 2/128) | 59 | 47 | 1 |
| (undated) | `bytegans` | Project One | 10 | 5 | 0 |
| (undated) | `bytegans__project-five-jkmzy-9cxsr` |  | 0 | 0 | 0 |
| (undated) | `bytegans__project-four-yjynj-ws7nk` | Make it stand out. | 83 | 20 | 0 |
| (undated) | `bytegans__project-one-ephnc-nnkp3` | Make it stand out. | 26 | 20 | 0 |
| (undated) | `bytegans__project-six-sz8wl-72are` | It all begins with an idea. Maybe you want to launch a business. Maybe you want to turn a hobby into something more. Or maybe you have a creative project to share with the world. Whatever it is, the way you tell your story online can make all the difference. Don’t worry about sounding professional. Sound like you. There are over 1.5 billion websites out there, but your story is what’s going to separate this one from the rest. If you read the words back and don’t hear your own voice in your head, that’s a good sign you still have more work to do. | 48 | 36 | 0 |
| (undated) | `bytegans__project-three-8zgh7-6letk` | Whatever it is, the way you tell your story online can make all the difference. | 32 | 19 | 0 |
| (undated) | `bytegans__project-two-llrgk-mxcl9` | It all begins with an idea. Maybe you want to launch a business. Maybe you want to turn a hobby into something more. Or maybe you have a creative project to share with the world. Whatever it is, the way you tell your story online can make all the difference. Don’t worry about sounding professional. Sound like you. There are over 1.5 billion websites out there, but your story is what’s going to separate this one from the rest. If you read the words back and don’t hear your own voice in your head, that’s a good sign you still have more work to do. | 48 | 36 | 0 |
| (undated) | `drop` | The First Sparks of Artificial Creativity | 167 | 65 | 28 |
| (undated) | `journal` | Excellent. You have found the Hidden Journal. | 22 | 20 | 0 |
| (undated) | `news` | My robots recently began painting mysterious images that revealed themselves to be bitGANs. | 17 | 14 | 0 |
| (undated) | `superrarebitgans` | bitGANs, SuperRare, and Sets | 99 | 67 | 7 |
| 2021-08-01 | `news__the-reveal` | The Reveal | 107 | 64 | 0 |
| 2021-08-08 | `news__the-dance` | The Dance | 43 | 28 | 0 |
| 2021-08-13 | `news__the-noungans` | The nounGANs | 54 | 37 | 6 |
| 2021-08-15 | `news__the-prime-noungan` | The Prime | 62 | 43 | 0 |
| 2021-08-19 | `news__bitgans` | The bitGANs? | 75 | 47 | 0 |
| 2021-08-21 | `journal__page001` | 001 - The Hidden Journal | 62 | 42 | 0 |
| 2021-08-22 | `news__a-glitch` | A Glitch | 24 | 19 | 0 |
| 2021-08-23 | `news__more-glitches-1` | More Glitches | 19 | 18 | 0 |
| 2021-08-24 | `news__the-glitch-is-spreading` | The Glitch is Spreading | 24 | 21 | 0 |
| 2021-08-26 | `news__cryptovoxels` | CryptoVoxels | 31 | 27 | 0 |
| 2021-08-27 | `news__the-drop` | The Drop | 75 | 49 | 1 |
| 2021-08-27 | `news__they-materialized` | They Have Materialized | 16 | 14 | 0 |
| 2021-08-29 | `news__hidden-patterns` | Hidden Patterns | 25 | 23 | 0 |
| 2021-08-31 | `journal__page002` | 002 - Hidden Types | 26 | 13 | 15 |
| 2021-08-31 | `news__hidden-types` | Hidden Types | 184 | 88 | 39 |
| 2021-09-01 | `news__poap-01` | POAP? | 60 | 41 | 1 |
| 2021-09-03 | `news__new-bitgans` | New bitGANs | 6 | 6 | 0 |
| 2021-09-04 | `journal__page003` | 003 - Fibonacci skullGANs | 70 | 50 | 12 |
| 2021-09-05 | `news__first-look` | First Look | 12 | 11 | 0 |
| 2021-09-06 | `news__dad-what-is-this` | Dad, what is this? | 32 | 20 | 0 |
| 2021-09-08 | `journal__page004` | 004 - Prime skullGANs | 25 | 21 | 0 |
| 2021-09-08 | `news__the-escape` | The Escape | 73 | 45 | 0 |
| 2021-09-09 | `news__64-more-renegades` | 64 More Renegades | 57 | 38 | 2 |
| 2021-09-11 | `journal__page005` | 005 - squareGANs | 100 | 60 | 0 |
| 2021-09-12 | `journal__page006` | 006 - Odd skullGANs | 14 | 14 | 0 |
| 2021-09-17 | `news__text-messages` | Text Messages? | 139 | 66 | 0 |
| 2021-09-21 | `news__materialization` | Materialization | 29 | 24 | 0 |
| 2021-09-24 | `news__20219291202` | 2021.9.29 12:02 | 46 | 34 | 0 |
| 2021-09-25 | `journal__page007` | 007 - Even bitGANs | 21 | 18 | 0 |
| 2021-09-25 | `news__anotherpainting` | Another Painting | 10 | 8 | 0 |
| 2021-09-28 | `news__ritual-sacrifice` | Ritual Sacrifice? | 25 | 21 | 0 |
| 2021-09-28 | `news__the-break-out` | & Another Break Out | 34 | 31 | 0 |
| 2021-09-28 | `news__the-puzzle` | The Puzzle (solved) | 22 | 17 | 1 |
| 2021-09-29 | `journal__page008` | 008 - How to Forge bitGANs | 249 | 116 | 12 |
| 2021-10-02 | `news__the-cryptographer` | The Cryptographer | 207 | 107 | 26 |
| 2021-10-08 | `news__burntganseth` | burntgans.eth | 17 | 15 | 0 |
| 2021-10-13 | `news__new-bitgallery` | New bitGallery | 3 | 3 | 0 |
| 2021-10-20 | `news__superrare-bitgans` | A Brief History of bitGANs | 311 | 141 | 18 |
| 2021-10-25 | `news__a-new-portal` | A New Portal (Solved?) | 40 | 33 | 4 |
| 2021-10-27 | `news__testnet` | Testnet? | 16 | 13 | 0 |
| 2021-10-28 | `news__testnet-auctions` | Testnet Auctions | 103 | 48 | 10 |
| 2021-11-03 | `news__the-portal-responds` | The Portal Responds | 8 | 8 | 3 |
| 2021-11-08 | `journal__page009` | 009 - More Forged bitGANs | 35 | 30 | 0 |
| 2021-11-12 | `news__the-ides-are-here` | The IDES are Here | 104 | 64 | 5 |
| 2021-11-13 | `news__the-final-clue` | The Final Clue? | 58 | 46 | 18 |
| 2021-11-13 | `news__what-have-we-done` | What have we done? | 66 | 46 | 1 |
| 2021-11-14 | `journal__page010` | 010 - podGANs | 1380 | 625 | 6 |
| 2021-11-19 | `news__more-pods` | more pods | 11 | 6 | 0 |
| 2021-11-19 | `news__pods` | pods | 73 | 44 | 2 |
| 2021-11-24 | `news__podgallery-puzzle` | podGALLERY puzzle? | 51 | 37 | 4 |
| 2021-11-28 | `journal__page011` | 011 - Decentralized Intelligence | 137 | 85 | 5 |
| 2021-12-03 | `news__an-invasion` | An Invasion? | 14 | 14 | 0 |
| 2021-12-03 | `news__podgallery-solution` | podGallery Solution | 112 | 53 | 4 |
| 2021-12-04 | `news__the-podgan-invasion` | An Invasion! | 154 | 88 | 9 |
| 2021-12-11 | `journal__page012` | 012 - cryptoGANs | 32 | 26 | 0 |
| 2021-12-18 | `news__the-winter-solstice` | The Winter Solstice | 60 | 38 | 0 |
| 2021-12-21 | `news__sun-dancing` | Sun Dancing | 145 | 87 | 1 |
| 2021-12-27 | `journal__page013` | 013 - Seasonal Patterns | 50 | 35 | 0 |
| 2022-01-01 | `journal__page014` | 014 - BIP39 and bitGANs | 64 | 39 | 1 |
| 2022-01-01 | `journal__page015` | 015 - bitGANs on arweave | 27 | 9 | 0 |
| 2022-01-09 | `news__pattern-search` | The Pattern Search | 274 | 134 | 0 |
| 2022-01-14 | `news__pattern-revealed` | Pattern Revealed | 70 | 41 | 6 |
| 2022-01-14 | `news__the-testgans-need-our-help` | The testGANs need our Help | 20 | 15 | 1 |
| 2022-01-25 | `news__22-for-2022` | 32 for 22 for 2022 | 116 | 76 | 2 |
| 2022-02-05 | `news__cryptoartgans` | cryptoArtGANs | 45 | 34 | 1 |

### 4.1 Notes on the pages that matter

**`news/the-reveal` (1 Aug 2021)** through **`news/the-glitch-is-spreading` (24 Aug 2021)** are
the origin-story posts. Three of them establish the word "glitch" as project vocabulary:
`news/a-glitch` ("But this one appears to have a glitch?"), `news/more-glitches-1` ("After the
first glitched out GAN appeared a couple days ago, many more have begun to form"), and
`news/the-glitch-is-spreading` ("The glitch from the animations has spread into my studio").
None of the three contains eleven of anything, and none has any emphasis markup. Their value is
that they fix "glitch" as a thing an artwork *is*, not a thing a word *is* — which supports
reading "11 glitches" as eleven artworks rather than eleven corrupted words.

**`news/hidden-types` (31 Aug 2021)** is the densest early puzzle post, with 39 emphasis tags,
and it is where the house style of numeric puzzles is set: Fibonacci frame positions, the
`[01123581321345589]` password, and the five Hidden Type classes. It also plants a deliberate
delayed fuse — "These are not 'Hidden Typos', but part of the puzzle that will only be solvable
after some more drops. Hold on to this info for later" — referring to `(32/128)` being Type 32
and `(119/128)` being the only odd Type E. Anyone hunting eleven anomalous artworks should note
that Pindar has form for flagging metadata irregularities years in advance.

**`news/the-puzzle` (28 Sep 2021)** is the roboGAN solution page. Analysed in full at §3.1.

**`news/the-final-clue` (13 Nov 2021)** is instructive about how he delivers clues: the clue is
never in the prose. Here it is inside a rendered scene — "He looked closely into the eyes on the
back wall… and was transported into an alternate dimension. The alternate dimensions contained
a hidden message." The eighteen emphasis tags on that page are all community members' names and
the numbers 3, 9, 27, 81, which is his other habit — bolding people and integers, never seed
words.

**`news/podgallery-solution` (3 Dec 2021)** is the SLIP-39/BIP-39 re-indexing puzzle. Its four
`<h3>` stage headings are the clearest statement anywhere of how he thinks a solution decomposes:
find the anomalies, recognise the pattern that selects twelve of them, read the attribute, fix
the encoding.

**`journal/page013` (27 Dec 2021)** and **`journal/page014` (1 Jan 2022)** are the Solstice clue
pages. Page 014 contains the only `<em>` on either — the single word *distraction*, italicised
inside "But I am beginning to wonder if their essence is just a *distraction*." That italic is
load-bearing: it is the tell that the Essence property, the one place everybody looks, was the
wrong place. It is the closest thing in the archive to a hidden hint delivered by typography,
and it is worth weighing against the fact that the Act 1 Puzzle page contains no italic at all.

**`news/pattern-search` (9 Jan 2022)** is a 6,276-character catalogue of 34 numbered community
theories, and it is the richest single page in the archive for method. Item 26 is directly
relevant: "The words 'potato' 'elite' 'book' and 'moon' were found in two of the cryptoArtGANs."
Item 33 notes four BIP-39 words skipped in the sequential ordering. Item 3 restates the clue
"Their essence is just a distraction, they hide their true BIP39 nature elsewhere." The page
carries zero emphasis tags, which is itself telling: when he is summarising other people's
work he does not bold anything.

**`news/pattern-revealed` (14 Jan 2022)** is the Solstice solution. Analysed at §3.2. It is the
only page before the cutoff whose bold text is a seed phrase.

**`news/22-for-2022` (25 Jan 2022)** is an auction announcement with nothing puzzle-related.

**`news/cryptoartgans` (5 Feb 2022)** is the Act 1 Puzzle page. Analysed at §2.

Three pages before the cutoff do yield exactly eleven BIP-39 words under some rule, and all
three are coincidences of length rather than design. `news/first-look` (5 Sep 2021) is a
368-character post whose body has 12 BIP-39 tokens and 11 distinct. `news/more-pods`
(19 Nov 2021) is a caption-only gallery post with 11 tokens, six of which are the word "view"
repeated from the "View fullsize" gallery captions. `news/the-escape` (8 Sep 2021) yields 11
BIP-39 words if you read only its headings. I derived and tested all three anyway; all fail.

### 4.2 Pages after the cutoff that comment in hindsight

**`journal/page016`, "016 - Act 1 Finale" (7 Feb 2022)**, two days after the puzzle was posted,
closes Act 1 and gives the drop census: 64 nounGANs, 128 skullGANs, 64 roboGANs, 12 SuperRare
bitGANs, 16 spookyGANs, 1024 podGANs, 92 solGANs, 32 testGANs, 112 ghostGANs, 512 bitGANs in
total, of which 64 are "Act 1 - Review" pieces released slowly over the following months. It
never mentions the puzzle.

**`journal/page015`, "015 - bitGANs on arweave" (1 Jan 2022)** gives the permanent-web archive
of all Act 1 properties as a CSV at `https://arweave.net/UwiYOGcO2pp-uRcyLnFxgYeSAwBzKSl91Nzz_wpypuw`,
plus per-drop ZIPs with the ranges spelled out: nounGANs 0–63, skullGANs 1–128, roboGANs
129–192, spookyGAN 193–208, solGANs 209–300, ghostGANs 301–444, Act 1 Review 445–512. Given
that all three solved puzzles drew their words from properties, this CSV is the natural
haystack for the eleven glitches.

**`news/1000-eth` (15 Feb 2022)**, **`news/wireless-podgans` (27 Feb 2022)** and
**`news/act-2-the-collaboration` (10 Mar 2022)** carry on as if the Act 1 Puzzle had not been
set. **`news/the-final-bitgan-drop` (23 Aug 2022)** describes the Act 1 - Review drop and names
the Quantum glitchGAN (512/512) as the last piece. **`news/the-bytegan-puzzle` (3 Mar 2023)**
lists four solved puzzles by URL and the Act 1 Puzzle is not among them, confirming it was
still open at that date. No page anywhere in the archive, at any date, refers to the Act 1
Puzzle again.

---

## 5. What I would do next

1. **Retrieve `clue_subtle_888.gif`, `clue_solution.gif` and `csv_list.png` from the roboGAN
   solution page, and `roboGANproperties.csv` from Google Drive.** These are the only surviving
   worked example of a Pindar clue-to-answer mapping and the local archive does not have them.
   Until they are read, "the roboGAN method" is being cited from a one-sentence summary rather
   than from the artefact.
2. **Pull `gitm_3__1001.gif`** — the Act 1 Puzzle page's unshown social-card image — and
   examine its frames and its embedded metadata. The byteGAN puzzle was solved twice over by
   exactly that technique, once from Photoshop layers in a GIF and once from faint per-frame
   inscriptions.
3. **Stop looking for the eleven words in prose.** Three for three, his seed words have come
   from artwork metadata, ordered by mint order. The phrase "11 glitches" is a pointer to
   eleven artworks; the search should be for the property that eleven glitch-related pieces
   share and the rest do not.
4. Treat the sets in `theories/puzzle_pages_sets.json` as closed on the five standard paths.
   If someone wants to reopen them, the only untested dimension is the derivation path.