1K Gallery

Overview

Pop Genius is one of the two collections that open 1K Gallery, both released on 2 October 2026. The series presents 1,000 original pop-art portraits inspired by Albert Einstein, composed from 40 trait layers across six categories: background, clothing, face, hair, moustache and tongue.

Every piece is a distinct combination. The trait layers were generated with the assistance of artificial intelligence, then reviewed, corrected and approved one by one before any combination existed. The final works were not produced as random prompt variations: each image is a deterministic composition of approved layers, selected by a seeded algorithm whose seed and script hash are published below.

The source gesture

The collection derives from a photograph taken on 14 March 1951, at a party held by the Institute for Advanced Study in Princeton for Albert Einstein's seventy-second birthday. After a full day of being asked to smile, Einstein stuck his tongue out at the United Press photographer Arthur Sasse. The picture went on to become as widely recognized as the equation that made him famous.

Pop Genius does not reproduce that photograph. It takes the gesture — the collision between a serious mind and an unserious face — as the starting point for a drawn, generative series in a different visual language.

Core specifications

Creation and curation process

1. The 40 artistic layers were generated with the assistance of artificial intelligence, then reviewed and approved one by one across the six trait categories.
2. Layers rejected in review were removed from the pool before any combination was generated, so that no discarded variant could reach the final collection.
3. The system calculated the 62,500 mathematically possible combinations from the approved variants.
4. Exactly 1,000 combinations were selected by a seeded algorithm: exact counts per value, grail reservation, per-category shuffling, and de-duplication by swapping.
5. All combinations underwent automated validation and visual review.
6. Approved images were recomposed from the master files at their native 2,000 × 2,000-pixel resolution.
7. Final files were checked for quantity, dimensions, sequence, metadata correspondence, and cryptographic uniqueness.

Trait distribution

The percentages below represent the exact frequency of each trait within the 1,000-NFT supply.

T1 — Background

T2 — Clothing

T3 — Face

T4 — Hair

T5 — Moustache

T6 — Tongue

Missing traits in the metadata

There are none. Unlike the gallery's first collection, where 40% of the items carried no necklace and the corresponding metadata field was deliberately left empty, Pop Genius has no optional category. All six categories carry a value in all 1,000 items.

This has a direct consequence for rarity: in the first collection the absence of a trait was itself the most frequent value in its category, and pulled the rarity of those items down. Here every item scores on all six categories, so no artwork is penalised for a field that was never filled.

How rarity is calculated

The collection does not include a Rarity Tier trait and does not pre-classify NFTs as Common, Rare, or Legendary. Rarity is a mathematical consequence of trait frequencies and the specific combination present in each NFT.

The score published with the collection is the sum, across the six categories, of the inverse frequency of the value each item carries:

Where count is the number of items in the whole collection carrying that same value — the figures in the distribution tables above. A value present in 30 items contributes 33.33 to the score; one present in 320 contributes 3.13. The rarer the value, the more it adds.

Worked example — token #1

Token #1 carries Mint background, Argyle clothing, Ochre face, White hair, White moustache and Pink tongue:

Ranks run from 1, the rarest item in the collection, to 1,000, the most common. Anyone can reproduce both the score and the rank from the public metadata alone, without trusting this document.

Grails

Ten items carry the rarest value in four categories at once. They were reserved by the allocation algorithm rather than emerging by chance, and their existence is stated in the collection's public description on every marketplace.

Rarity across Token IDs

In the gallery's inaugural collection the artwork-to-token assignment was not shuffled: numbering followed the order of generation, which produced a measurable rarity gradient across token IDs. Items became steadily more common towards the end of the numbering, and the cause was traced to the clustering of the no-necklace items in the final blocks. That gradient was disclosed in full on the first collection's technical sheet, together with a commitment:

Pop Genius is that collection, and the commitment is honoured: the allocation is shuffled per category and de-duplicated by swapping, driven by seed 20260915. Both the seed and the SHA-256 of the script that consumed it are published in this sheet, so the assignment can be reproduced independently and shown not to have been arranged after the fact.

Why a gradient is not possible here

The argument is structural before it is statistical. The shuffle is the last step of the generator: the 1,000 rows are assembled, and only then reordered, before numbers are attached. There is no stage after it in which an ordering could reassert itself. A systematic gradient across token IDs is therefore impossible by construction, not merely unlikely.

That distinction matters, because a collection this size will always show some variation between blocks of 100 — that is what sampling noise looks like. Reading such variation as a finding, by choosing which range to test after seeing the data, produces impressive-looking results that mean nothing. The mechanism is what rules the gradient out; the measurement below only confirms the code did what it says.

Measured distribution by ID range

For each block of 100 token IDs: the average rarity rank, and how many of the collection's 100 rarest items fall inside it. Rank 1 is the rarest item in the collection and rank 1,000 the most common, so a block with no bias averages near 500.

The averages run from 466 to 535 — a spread of 69 positions around a collection mean of 501. From the first block to the last the difference is −38: the end of the numbering is marginally rarer than the beginning, which is the opposite of a gradient and the signature of noise rather than order.

The contrast with the inaugural collection is the point. There the block averages ran from 410 to 708, a swing of nearly 300 positions that climbed steadily through the final third, and the 100 rarest items thinned out from 25 in one early block to a single one in the last. Here the rarest hundred are spread across every block, and no block is a better or worse place to hold a token than any other.

Items reserved for the gallery

100 items, the same number as the inaugural collection. They are minted from the gallery's own wallet in the allowlist stage, before the public opening.

The token numbers are not fixed in advance. The contract issues IDs in the order the transactions land, so the gallery's block is whatever the mint produces on the night — and it is published here, in full, once it exists.

File integrity and provenance

Each of the 1,000 images has a unique trait combination and a distinct SHA-256 hash. The values below document the version prepared before publication without revealing the individual manifest or the association between images and Token IDs.

The provenance hash is computed over the bytes of the image manifest exactly as delivered: 73,893 bytes, LF line endings, no byte-order mark. Reopening and saving that file in a text editor would change its hash and invalidate the published provenance, which is why the delivered file is never edited.

The complete individual image-hash manifest will be disclosed only after the reveal.

Images and metadata

Image optimisation

• All images are 2,000 × 2,000 pixels with a square composition.
• Delivery is JPEG at quality 92, progressive, with an embedded sRGB profile and no EXIF data.
• Chroma subsampling is 4:4:4, not the usual 4:2:0. The artwork has saturated flat areas and hard outlines, exactly where reduced chroma resolution shows as blurred colour edges; full chroma costs file size and keeps the edges clean.
• EXIF is stripped so that no camera, software or location metadata travels with the artwork.
• Each file is approximately 923 KB; the complete set is 0.95 GB.

Metadata

• The metadata contains 1,000 records encoded in UTF-8, one for each NFT.
• The metadata uses six public categories: Background, Clothing, Face, Hair, Mustache, and Tongue.
• No category is ever empty: every record carries a value in all six.

Filename correspondence

• The sequence 1.jpg through 1000.jpg corresponds directly to Pop Genius #1 through Pop Genius #1000.
• The tokenID field in the metadata matches the numeric part of the filename, with no offset and no zero padding.
• Because the contract concatenates the token number onto the base URI, the base URI is written with a trailing slash; the collection URI, which points at a single file, is written without one.

Contract, storage and license

The values below are verified directly from the contract and from permanent storage. Those marked as published on deployment only exist once the collection is deployed, and are filled in here before the first mint.

Still to publish

Everything in this sheet is final except the values below. All of them are filled in before the first mint:

• Token numbers of the 100 items held by the gallery, after the mint
• Contract address of the deployed clone
• Deployment block and date