Orkhan Mammadov

Under the GAN

Craft proof: an immersive archive of early AI art, from the first machine dreams to autonomous creativity.

Client
Own product, for the Jediwolf collection
Role
Product owner
Design
Year
2026
Discipline
Product Design
Web
AI / ML Interfaces
Prototyping
Under the GAN landing page: the collection scattered as a cloud of works behind the wordmark, with the line Machines began in silence. This is the decade they learned to see.
[ THE PROBLEM ]

A private collection of early AI art, 198 works by 61 artists from DeepDream in 2015 to autonomous systems in 2026, needed a home that could be two things at once: an immersive gallery worth entering, and an archive of record whose provenance, bios and curatorial framework hold up. Collection sites usually pick one. This one had to be walked, orbited and mapped, and still read like a catalogue.

[ GOAL ]

An immersive archive of early AI art, from the first machine dreams to autonomous creativity: the decade machines learned to see, kept as a collection you can enter.

[ SOLUTION ]

1. Ten ways of seeing the same 198 works: walk them in a museum, orbit them as a field or a globe, map them as a network, or read them as a grid, a timeline and an index. 2. A curator CMS underneath: roster-driven publishing, a dashboard of what needs attention, and a nightly on-chain sync that imports new tokens as pending and never auto-publishes. 3. Provenance as content: every work carries its Ethereum contract and token, the work page tells its chain history as a story, and the network view draws the collection wallet's graph.

Ten ways of seeing

[ TEN WAYS OF SEEING ]

The same works, encountered through different structures: as an archive, a museum, a timeline, a field, a network. Four of the views run on three.js; the rest are plain DOM, so the collection always renders.

The Museum view: a first-person three.js gallery, Chapter 01 Awakening 2015 to 2016, works hung on both walls with reflections on the floor
Enter the collection in first person

The landing

[ THE LANDING ]

A console boots the archive, the collection scatters behind the wordmark, and the page ends on the choice of how to see it.

The console boot sequence that opens Under the GAN: loading archive, 198 works, 62 artists, eras calibrating, an Enter button
A console boot: the archive loads in front of you
The Ten Ways of Seeing chooser on the landing page: Latent Space, Museum, Index, List, Horizontal, Grid Gallery, Canvas, Timeline, Sphere, Network
Ten ways of seeing, chosen on the landing page

A work, with its history

[ A WORK, WITH ITS HISTORY ]

Each work page carries the piece, its model and facets, and a provenance story read from Ethereum: every hand it changed is written to the chain.

A work page: the piece centred with its media, model and facets
A work page: the piece, its model and its facets
A provenance story beat on a work page: Since mint, this token has passed through 2 addresses, every hand it changed is written to the chain
Every hand it changed is written to the chain
The On-Chain page: the collection wallet read live from Ethereum, with balance, collections, transactions, token transfers and curated works
The collection wallet, read live from Ethereum

Artists and the editorial layer

[ ARTISTS AND THE EDITORIAL LAYER ]

Sixty-one artists with fact-checked bios, an About that tells the decade in four chapters, a field guide to the models, and exhibitions hung from the collection.

The Artists index of Under the GAN
Sixty-one artists, each with a fact-checked bio
An artist page: Anna Ridler, 2 works in the collection, British, Forming, 2019
An artist page, led by the signature series
The About page prologue: How Machines Learned to See
How machines learned to see, in four chapters
The About page chapters: It begins in silence, the curatorial framework in four chapters
It begins in silence
The Field Guide: a neural network diagram, signal flowing left to right layer by layer, with the chapter navigation from Neural Nets to The Artists
A field guide to the models behind the collection
The Learning the Face exhibition: a dense hang of faces across the decade
Learning the Face, a hang of faces across the decade

Summary

[ SUMMARY ]

Live at underthegan.art since July 2026: a Vite, React and three.js front end over a Payload CMS on a single server, every route prerendered with its own metadata, analytics self-hosted, and an AI-assisted build throughout. Curatorial direction by Kate Vass; infrastructure by Novexing.

198
works in the collection
61
artists, each with a bio
4
curatorial eras, 2015 to 2026
10
ways of seeing
The Sphere view: works and artist portraits wrapped onto a rotating globe