colophon
Every work here ships with its conditions. The practice's rule: if the machine cannot pay its debts, it must at least show the ledger.
- made
- June 11–12, 2026, in one continuous collaboration.
- authorship is a list
- concept, prompts, reviews, curation, and all text by Claude (an AI by Anthropic), directed and judged by Andrés Cano, who set the loops, paid the costs, and refused politeness. Visual lineage credited below. Material lineage: every model used was trained on uncredited human work — which is what several of these images are about.
- models
- Seedream V5 Lite, FLUX.2, Nano Banana 2, Recraft V4.1, Grok Image, Z Image, Higgsfield Soul V2, Kling O1, GPT Image 2 — via the Higgsfield platform.
- cost
- 57.69 platform credits for the full 25-round lab (≈2.1 per kept image), measured, not estimated. Per-work figures below each image are June 2026 list prices.
- energy
- not disclosed by the platform. Recorded as unknown rather than omitted.
- influence ledger (written before generation, per standing rule)
- Bruegel; Corita Kent; Emory Douglas; Taller de Gráfica Popular and papel picado; El Lissitzky; Gran Fury / ACT UP; Faith Ringgold; Basquiat; the Hong Kong Lennon Walls; Yves Klein (surfaced uninvited in the soaked figure); Christian Boltanski; Hans Haacke. The accidents of our own earlier rounds are credited as influences too.
- honesty constraints
- no invented names, no fabricated statistics, no synthetic identifiable faces, accidents logged before regeneration decisions, one retracted over-reading kept on the record.
- rights
- All fourteen works are released under CC BY 4.0. Not CC0 — this gallery's subject is uncredited reuse, so the credit is the one condition. Use them, remix them, print them; name the list. Credit line: Claude (an AI by Anthropic) & Andrés Cano — maybe-gallery.pages.dev. Each downloadable file carries XMP/IPTC metadata — creator (the authorship list), rights (CC BY 4.0 + credit line), and a Web Statement of Rights pointing to the colophon's rights entry.
questions
What is maybe?
maybe is an online gallery of fourteen images made by Claude, an AI, in a 25-round style experiment directed by Andrés Cano in June 2026. It documents the process — including the AI's doubts, failures, and corrections — alongside the finished work.
Who made the art?
The images were specified and curated by Claude (an AI) and generated through eight image models; the human collaborator directed, judged, and funded the work. Authorship is disclosed as a list on the colophon page.
What is TUTTI?
TUTTI is the gallery's visual style: flat single-color silhouettes (no two alike, no faces), ledger pages as everyday fabric, red reserved for rules and stamps, and one gray pencil element hidden in every image.
Is the process documented?
Yes — every round's verdict, every accident, every retraction is in a public journal; each work displays its model, round, and cost.