Brett Amory
Artist · Educator · Builds with AI
I'm an interdisciplinary artist based in Oakland. I've taught studio art at Stanford since 2019 and at Cañada College since 2024. I've used AI as a working tool and collaborator every day since generative image tools arrived in 2021.
I'm not a programmer. I build software by directing AI coding agents: I design the system and decide what it should do, and the agents write the code. Everything on this page was built that way.
Unself Oracle
2024 – present · daily videoOne episode: a real news story, rewritten by the language, spoken by UNSELF.
UNSELF, a flat white mask no face detector can find.
A daily video in which a masked figure reads a news story rewritten by the language, in a cloned voice I tuned over fourteen rounds of tests. The lip-sync tools I tried couldn't find a face in the blank mask, so I typed its coordinates in by hand.
Every video ends with one sentence in the exact mode, glyphs and English together.
Concept, mask, voice and training data. I designed the whole pipeline and built it by directing AI coding agents: a trained language model writes the script, a cloned voice reads it, and Wav2Lip lip-syncs the mask.
The same day, told twice: a news story as published, and what the Oracle made of it.
z0glyphic 913
2022 – present · constructed languageWhen generative image tools arrived in 2021, I used them the way I had used the copy machines at Kinko's in the late nineties, looking for where they break. The models kept producing invented text that no one had written. I collected it, and it became z0glyphic 913.
It now has 142 base signs, a grammar, more than 12,000 words, a drawn typeface, a spoken form made of clicks and my own recorded voice, and a codec that turns English into glyphs and back. Automated jobs coin new words every four hours, and every change has to pass a fixed set of test sentences or it gets thrown out.
Everything: the glyphs, the grammar, the dictionary, the translator, the trained models and the public repository, which a script rebuilds from the language's own files so it can't fall out of date.
Word card: eye. One of a series of prints pairing a word with a painting.
The public repository, language.brettamory.com.
The translator: English to glyphs and back.
The marks the language is built from, each with one locked meaning.
Surveil Thyself
2026 – present · daily system
Daily posters.
One cycle, five stations.
This system reads my phone photographs and daily digital traces the way an advertising company would. It has sorted my 41,312-image archive into thirteen bodies of work I call Spells.
Every morning at 07:45 the system reads the previous day through z0glyphic 913 and prints a plate from it: one of my photographs, a generated ground, a line from the day's reading, and the day's glyph signature along the bottom. Eight layouts render every day from the same material.
Designed the trackers, the reading, the poster layouts and the grouping into Spells, and built them as scheduled jobs that run every morning without me.
Paintings
Silence Is Louder Than Sound · 2025The paintings that come out of this are still made by hand, in a fixed order of layers: collaged photographs, an acrylic wash, a screen-printed glyph layer, a layer from a small image model trained only on that body of work's pictures, and oil on top.
My 2025 solo exhibition at The Space Program in San Francisco, Silence Is Louder Than Sound, was the first show made this way, and the media theorist Shane Denson wrote about it in "The Aesthetic Halting Problem."
Earlier work has shown at Jonathan Levine Gallery, Lazarides, the de Young, the Fort Wayne Museum of Art, and the National Portrait Gallery in London, where it received a BP Award in 2016. In 2024 I received a Pollock-Krasner Foundation award.
Installation view, The Space Program, San Francisco, 2025.
KAKO, 2025. Ink on canvas, 72 × 96 in.
Archive of Failed CAPTCHA Attempts, 2025. Ink on paper and canvas, 36 × 36 in.
Diagram for Remembering Nothing, 2025. Ink and oil on canvas, 72 × 84 in.
Studio systems
2025 – present · tools I built for my own studio
Engine Room, the system's own switchboard: one loop through Capture, Language, Produce, Spells and Show.
The Prompt Book: a wiki built from 759 image-generation prompts, distilled into templates, rules and failure modes.
Around all of that sit the tools that keep a working studio running: a dashboard I call Mission Control, a design system for keeping visual work consistent, image tools that resize and fit pictures for print, portals and Instagram, a generator that builds layered Photoshop files, and several websites, including the language's public repository and a password-gated learning site with a twelve-part audio course.
Explainer reel, made entirely in code: motion graphics rendered with Remotion.
Teaching
Stanford since 2019 · Cañada College since 2024I've taught Painting 1 and 2, Drawing 1 and 2, Color, Photoshop and Painting, and AI as a Creative Partner, and I received Stanford's Centennial Teaching Award in 2020.
What I bring to students is less a single technique than a working method. I show them how to treat these tools as material: train them on your own pictures, find where they fail, build the tool you need when it doesn't exist, and keep a hand on the final object.
Talks: AGI House (2024, 2025), Commonwealth Club (2024, 2025), Circuit Launch (2026).
An Albers color-study web app that works from students' own photographs, and a student-editable Painting II reference guide on Canvas at Stanford.