artist, he/him 🇨🇦
Joined February 2012
- Tweets17.7K
- Following1.7K
- Followers62.2K
- Likes36.3K
Pinned Tweet
I've just published "Windfoil: Closed-Form Coverage for Real-Time and Differentiable Vector Graphics" —
📎 Paper: arxiv.org/abs/2610.02468
🔧 Code: github.com/texel-org/windfoi…
It is substantially faster than current differentiable renderers, and runs in WebGPU (browser + server).
Matt DesLauriers retweeted
Revisited and optimized this one made back in 2017.
Inspired by @RijpmaJohan 's abstract stop-motion works, I set up a feedback loop between a texture and the main graphics context.
Excited to share that I'll be exhibiting "Partial Exposure," a new series of works at Paris Photo, Nov 11 - 15th at the Grand Palais as part of the @automataisart group show with @AmeesiaMarold, shown alongside other works by @harmvddorpel @100o111 @isthisanart_ @solienne_ai.
A selection of exposures will be preserved as framed archival-quality prints, 35mm film transparencies, and bespoke View-Master reels, creating physical mementos of scenes that feel remembered, although never experienced.
This is based on my PhD research and the differentiable renderer that I've been building these last few months, including a pure-WebGPU CLIP encoder that is optimised for real-time and interactive use.
You can read the Windfoil paper here:
arxiv.org/abs/2610.02468
pretty wild - our models are sitting just a couple decimal points away from an evil AI overlord mode. 🫠
if you have a qwen3.8 27b around, try steering L34 N3838 +0.75, now you have evil qwen 27b
for extra evil add L38 N17180 +0.75
same recipe as @_lyraaaa_: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖ at the down_proj input, every token
Matt DesLauriers retweeted
I am excited to present Scenography @ParisPhotoFair in just two weeks!
A group show by @automataisart featuring new works from...
@harmvddorpel
@mattdesl
@100o111
@isthis
@solienne_ai
November 11-15th, 2026
Grand Palais, Paris
Booth F08
Learn more and explore the works ↓
A cut of ‘Generative’, a documentary on the history of generative & computer art, will be screened today in Marfa at 3.30 PM, at the Crowley Theatre.
@Generative_film
Matt DesLauriers retweeted
a new kind of entropy-based cellular automata
"decompressive cellular automata"
here, each pixel is 128 bits of data that is repeatedly decompressed with brotli-like algorithm
in addition neighboring pixels may swap bits (0..1 ramp left-to-right) or median filter bit planes in the image (0..1 ramp top-to-bottom)
Matt DesLauriers retweeted
Degrees of Disruption opened today at the V&A.
Two editions of Meridian by @mattdesl (#870 and #686) are now in the museum's permanent collection, and one is on display.
The series is documented in the Meridian book, published by Vetro: vetroeditions.com/products/m…
Meridian #870 is currently on display at the Victoria & Albert Museum in London as part of their new show, Degrees of Disruption: Vera Molnar and Generative Art. 🤗🎉
The exhibit runs until May 2027, and highlights the work of Vera Molnár alongside several contemporary artists.
Pictured below, works by @REAS @williamapan @zachlieberman @Licia_He, alongside Vera’s.
pretty wild display of code + vector graphic capabilities coming out of this recent Google model.
Gemini 4 Argon is the first model where It felt like an image, video and language model all at once.
Everything in this thread is made with code. @GoogleAIStudio
More examples + the full code for all 534 paintings below ⬇️
Matt DesLauriers retweeted
it's been too long since I worked on anything 3D! gosh, I missed @threejs
here's our interactive 3d magazine, made to showcase the amazing art of @agusegui and celebrate the official launch of Paper Mono!
see it live: paper.design/mono
more about this absolute gem of a font: teapawt.net/vladmoroz/status/21078…
team behind the website: @vladmoroz @alex_krasikau @agusegui
despite the agentic/AI era coming for our code, it feels like there's still room for a tool like 'canvas-sketch' for creative development. I've been making and testing a new tool in the last months, some notes:
- git SHA on export is still extremely useful; every output is tagged and correctly archived
- a single consistent template/format for every sketch, regardless of output medium and design considerations, keeps things very clean
- important that it works in both web (UI + realtime etc) and CLI (batch & server rendering)
- ability to run locally (eg: immediate access to GBs of models, data, images) feels crucial for R&D and creative experimentation
curious to hear what others think
github.com/mattdesl/canvas-s…
massive undertaking by @lerandomart mapping out lots of the history & space of generative art! 👏
Super fun, @WarnerTeddy used some of my evolutionary drawing system (Synthetic Gestures, ACM C&C ‘26) to make a bundle in Avian Visitors! 🐦
My publication with more info:
dl.acm.org/doi/10.1145/38037…
I made bird bundles to easily switch styles in your bird tracking collage and share with others avianvisitors.com/bundles