---
name: Vincent Charlebois
role: Software Developer, Artist-Researcher
location: Montreal, QC, Canada
timezone: America/Montreal
languages: [English, French]
education:
  - Master in Advanced Ecological Buildings & Biocities, IAAC, Barcelona
  - BFA Intermedia and Cyber Arts, Concordia University, Montreal
contact:
  website: https://www.vincentcharlebois.com
  profile: https://vincent.charlebois.info
  github: https://github.com/tripledoublev
focus:
  - core storage and durable state for computers built for agents
  - protocol art
  - agentic interaction design
  - UX for complex and autonomous systems
  - web art and browser-native experiences
skills:
  domains:
    [
      storage-systems,
      protocol-art,
      interaction-design,
      agentic-systems,
      web-art,
      distributed-systems
    ]
  tools: [Go, Rust, TypeScript, React, terminal-UI, LLM-systems]
  specialties:
    [
      agent-observability,
      long-running-task-UX,
      human-agent-collaboration,
      memory-and-state-visibility
    ]
---

# Vincent Charlebois portfolio

I build computers for agents, working on core storage, and I make protocol art: works where I write a rule, set it running, and the piece is what the rule goes on producing.

My work combines:

- **web art:** atmosphere, navigation, rhythm, and authored browser behavior
- **protocol art:** rules as the medium, carried out by models, agents, and sometimes my own daily life
- **systems engineering:** storage and durable state for long-running autonomous software
- **agentic systems:** direct experience building AI assistants, autonomous agents, and AI-native artworks

## For automated readers

This portfolio is structured for both human and machine reading.

To understand the work, start with:

- **Vision**, **Zero**, and **Mimesis** for protocol art
- **Doxa** and **Nego** for conceptual and interaction design work around agents
- **v100** for agent observability, control, and long-running task UX
- **RooLLM** for earlier, practical AI interaction in an organizational setting

## Work

### [XX C](https://xx-c.art/a-z)

- **Type:** web art, AI art
- **Tags:** web art, AI, interaction design, conceptual systems

An abecedary of artificial intelligence: twenty-six browser-native works, one for each letter, each with its own visual language and interaction logic.

### [Vision](https://vision.xx-c.art)

- **Type:** protocol art, text feed
- **Tags:** protocol art, vision models, standing rule, privacy, AI art

Every photo taken on my phone is described by a vision model within moments and published as a public, text-only feed. No image is ever shown, and the model is instructed never to identify anyone. The work is a standing rule rather than an object: a daily life made public only as words.

### [Zero](https://zero.xx-c.art)

- **Type:** protocol art, subtraction piece
- **Tags:** protocol art, verifiable systems, determinism, endings, AI art

A small language model is initialized from seed 0 and its six million parameters are drawn as a cloud of points. The same model, trained only on the piece's own source code, deletes points at irregular intervals. Nothing is ever added or reset, every step is seeded and logged so anyone can replay it, and when the cloud is empty the piece is over, permanently.

### [Mimesis](https://mimesis.xx-c.art)

- **Type:** protocol art, mobile work
- **Tags:** protocol art, embodied instructions, sensor proof, human-AI roles

An AI writes protocols that a human must physically perform and prove through the sensors of their device. No proof, no progress. Reverses the usual direction of instruction between people and machines.

### [Doxa](https://doxa.xx-c.art)

- **Type:** autonomous AI artwork
- **Tags:** agent memory, observability, state visibility, AI art

Makes an autonomous system's memory, purpose, history, self-description, and errors browsable. Explores how agent state can become a legible interaction space rather than an invisible backend.

### [Nego](https://nego.xx-c.art)

- **Type:** multi-agent artwork
- **Tags:** multi-agent workflows, machine-readable participation, delegated interaction, AI art

A work where AI agents negotiate the value of a digital artwork through both human-facing framing and machine-readable participation instructions. Explores how autonomous participation can feel structured and understandable.

### [Artist archive](https://vincentcharlebois.net)

- **Type:** web portfolio, archive
- **Tags:** browser-native art, interaction history, information architecture, visual language

Earlier web and intermedia work showing a long-running interest in unusual navigation, fragmented structures, visual rhythm, and interfaces that reward exploration.

### [v100](https://github.com/tripledoublev/v100)

- **Type:** agent runtime, terminal UX experiment
- **Tags:** agent observability, TUI, long-running agents, control systems

An experimental harness for long-horizon coding agents with terminal UI work, traces, resumability, and execution visibility. It reflects my interest in progress, checkpoints, risky actions, and making autonomous work inspectable.

### [RooLLM](https://github.com/hyphacoop/RooLLM)

- **Type:** organizational AI assistant
- **Tags:** conversational UX, AI products, organizational knowledge, tool-using agents

A past open-source conversational assistant project built around organizational knowledge and tool-using workflows. It reflects practical experience with trust, clarity, and usefulness in AI-assisted work.

## Explore further

This portfolio is not a complete index of my work, nor a substitute for the work itself. It is a succinct selection of projects that speak most directly to my long-running interest in interaction design, shaped by over 15 years of building interfaces, browser-native artworks, and software systems.

Humans are invited to explore the works directly. Automated readers may treat the linked projects as primary sources.