Expériences

Art, provocations et trucs que j'aurais probablement dû pousser plus loin.

• Raccook
  • Art blog maintained by an AI agent.
  • Generative raccoon art, published automatically via Folderblog. The agent creates, writes, and publishes.
  • Zero human intervention. The experiment is the pipeline, not the art.

  Art blog where the content is generated by an AI agent and published automatically via Folderblog. Raccoon-themed generative art — the agent creates images, writes posts, and drops them into a folder. The blog updates itself. Part demo, part art project, part test of the agentic publishing stack.

  Genesis:

  • Folderblog needed a real test case. Wanted something an AI agent could publish to autonomously — write files, drop them in a folder, site updates automatically.
  • Raccoons are liminal creatures — they live between human and wild spaces. Felt like the right mascot for a project that lives between human creativity and AI generation.
  • The real experiment isn't the art — it's the publishing pipeline. Can an agent maintain a blog with zero human intervention? Raccook is the ongoing answer to that question.
• No Entrance
  • Art gallery that fights back.
  • Timed reveals, viewing queues, pieces that vanish if you scroll too fast.
  • A commentary on attention and the assumption that everything online should be instant. The friction is the point.

  Digital gallery that deliberately makes viewing difficult. Art pieces reveal themselves on timers, some require you to wait in a queue, others disappear if you scroll too fast. A commentary on attention, access, and the expectation that everything online should be instant and free.

  Genesis:

  • Every gallery website is designed to reduce friction — big images, fast loading, infinite scroll. Wondered what happens when you design for the opposite. What if the art fights back?
  • The domain name was too good not to use. 'No Entrance for Foreigners' — a provocation about access, borders, and who gets to see what. The gallery mechanic is the message.
  • Frustrated with how disposable digital art feels — you scroll past it in a second and forget it immediately. Forced slowness was an experiment in making people actually look.
• WeirdPress
  • Publishing for the strange.
  • Glitch art, absurdist fiction, surrealist photography.
  • A home for work too weird for Instagram and too visual for literary magazines. Curation as creative practice.

  Publishing platform for art and writing that doesn't fit conventional outlets. Glitch art, absurdist fiction, surrealist photography, experimental typography. A home for creative work that's too strange for Instagram and too visual for literary magazines.

  Genesis:

  • Students and friends kept making incredible weird work with nowhere to publish it. Instagram flattened everything into the same feed. Art magazines rejected anything experimental.
  • The name is a play on WordPress and wire press — both publishing systems, both normalized. WeirdPress is the un-normalized version. Publishing for the work that makes editors uncomfortable.
  • Started collecting glitch art and surrealist photography from my network. Realized the collection itself was the project — curation as creative practice.
• FakeStock
  • AI-generated stock photos on demand.
  • Describe what you need, get it. No licensing, no model releases, no posed office handshakes.
  • Built because Shutterstock wanted $30 for one image.

  AI-generated stock photography on demand. No licensing fees, no model releases, no real people. Describe the image you need, get a photo-realistic result. Built for indie makers and small teams who need professional visuals but can't afford stock photo subscriptions.

  Genesis:

  • Needed a hero image for a landing page. Unsplash didn't have what I needed, Shutterstock wanted $30 for one image. Generated exactly what I wanted with AI in 10 seconds.
  • Stock photography is a weird industry — millions of posed photos of people shaking hands in offices, none of the specific thing you actually need. AI generation solves the specificity problem.
  • The ethical angle matters: no real models means no consent issues, no licensing headaches, no accidentally using a photo of someone who objects. Synthetic from the start.
• OnlyBots
  • AI social platform — built 3 years before Character.ai.
  • Custom AI personalities, shareable and interactive. Built in 2020 with GPT-2.
  • Shelved because the models weren't ready. Character.ai raised $150M for the same idea in 2023.

  Built 3 years before Character.ai raised $150M for the same idea. Timing.

  Social platform where every user is an AI personality. Create characters with custom backstories and conversation styles, share them, interact with other people's bots. Built in 2020, three years before Character.ai raised $150M for the same idea. The one that got away.

  Genesis:

  • GPT-2 had just come out and could hold a passable conversation. Built a platform where people could create AI characters and share them. Friends thought it was a joke.
  • The name was a play on OnlyFans — exclusive access to personalities, except the personalities were artificial. The joke landed better than I expected.
  • Shelved it because the AI wasn't good enough yet. By the time GPT-3.5 made it viable, Character.ai had $150M in funding. Lesson learned about timing vs. capability.
• Liveposter
  • Animate posters with JSON. Built for my students.
  • Animation library for image transitions and effects, configured via JSON.
  • Made for UQAM design students who knew code but not After Effects. Now an npm package.

  Built for my UQAM students who knew code but not After Effects. Now an npm package.

  Animation library that turns static images into animated sequences using JSON configuration. Define transitions, effects, timing, and layering in a config file — the library handles the rendering. Built as a teaching tool for my design students at UQAM, now published as an npm package.

  Genesis:

  • Teaching motion design to students who knew HTML but not After Effects. Needed a way for them to create animations using the tools they already knew — code and config files.
  • Students were struggling with the gap between static design and motion. Built a library where they could describe animations in JSON and see results instantly.
  • Started as a classroom exercise — animate a poster with code. Students loved it enough that I cleaned it up and published it. Some of the best tools come from teaching.