Post-Generation: Soup is Made of Water (2026)

Curated by Tina Tarighian, Post Generation was exhibited in October 2026 at TIAT, a San Francisco gallery dedicated to art and technology. Post Generation brings together three artists to make new work from an archive of test videos produced during the development of Google’s early Veo models. As generative video gets smoother and more convincing, the show looks back at what those first experiments left behind and asks what artists can find in the leftovers.

My contribution, Soup Is Made of Water, focuses on the text. Video models don’t write. They render letters as texture, the same way they render bark or wool, with no sense of what a sign is supposed to say. In the archive, this produces two kinds of failure: glyphs that look like an alphabet without belonging to one, and real, correctly spelled words strung together by association rather than meaning, like the phrase that gives the piece its title. Because the model redraws each frame without remembering the last, the letters also drift and melt as you watch.

I selected ten frames of this text and knit them as samplers on a hacked knitting machine. For centuries, the sampler was how girls learned their letters, stitching the alphabet one character at a time often without ever being taught to write with a pen. They mastered the form of writing without being allowed to use it to say anything. The model sits in a strange echo of that position: it produces the look of writing at enormous scale, with no one behind it trying to speak.

CBC Arts Logo (2026)

In January 2026, CBC Arts commissioned me to reinterpret their iconic gem logo as part of their monthly Canadian artist series. My concept looked back to CBC's earliest digital presence (the soft blues and whites of the network's debut online identity) as a way of thinking about what it feels like to become visible for the first time. There's something tender about those early institutional attempts at existing online: earnest, unpolished, almost handmade in their uncertainty. I produced the piece using a knitting machine and hand-dyed, made-in-Canada merino wool, with a backdrop of slight colour variation meant to evoke the pixel noise of early screens before they achieved the smoothness we now take for granted.

As part of this project, I also produced a behind-the-scenes Instagram Reel to walk through the creation process.

Ground Truth (2025-Ongoing)

Ground Truth is a series of wearable knits drawn from my experience as a data labeller, where I spent countless hours with the raw visual material that trains machine learning systems — training datasets, failed experiments, diagnostic visualizations that document AI's stuttering attempts at vision. Each piece translates the pixel densities, edge detections, and classification boundaries of machine vision sourced from computer vision papers into textile structures that respond to the body's movement, materializing the typically invisible and concealed labor that underpins digital systems.

Marine Learning (2025)

Marine Learning began with a question: how does a computer see water? Researching a vision model trained on underwater footage (one that required 125 frames per second to assemble meaning from motion and blur) revealed that machine vision, like human vision, is constructed, conditional, and prone to error. In response, I knitted a panel interpreting the model's early, faltering attempts to understand water, building meaning loop by loop the way the model built it frame by frame.

I then sent the panel to artist Alicia Wright, who responded with a crocheted frame derived from a 17th-century seaweed illustration sourced from the Internet Archive. A scientific drawing, once meant to capture objective truth, filtered through image tracing, screen glare, and hand. To finish, Alicia dyed the frame with water, pigment, and vinegar, binding it to the panel with SeaCel, a fiber spun from processed seaweed. Throughout, the boundaries between digital and physical kept dissolving: machine learning outputs became knit swatches, vector files became tangible objects. What began as a curiosity about artificial perception became a meditation on how meaning is made (imperfectly, in layers).

Gabbois x Backmarket (2024)

I worked as a production assistant for Gabbois' collaboration with BackMarket, where I was responsible for material sourcing, providing design consultation & prototyping, and overseeing production of a crochet balaclava made with recycled headphone wires.

Wingspan (2023)

Wingspan is an ongoing series of wearable pieces address the challenge of how you capture something made to move. Using flight-tracking diagrams and wing-beat analyses as source material for fair isle patterns, each piece translates scientific documentation into knitted form to reveal what any act of translation loses. Brushed mohair carries this well; it holds the ghost of touch, catches light differently when disturbed, remembers. The result is less representation than parallel system, a wearable trace of motion that acknowledges, from the start, that it cannot fully hold what it's after.