Thibaud Colas thibaudcolas

Overview template

Website

One-line positioning

Triage: JPEG XL image support

Assessed against selection criteria + product strategy v1.0 on 2026-07-07.

🏔️ JPEG XL image support

Add JPEG XL (JXL) encoding/decoding to Willow, then expose the format in Wagtail core for image uploads and renditions, enabling ~15% smaller file sizes than AVIF, reducing bandwidth and carbon footprint with minimal user effort.

Classification: Pillar work

Django Documentation Token Report

Total Documents: 656

Total Tokens: 10,801,338

LLMs.txt Files

File	Tokens

Evaluating Markdown support in LLMs

With explicit prompting for Markdown output (best case scenario).

Problem scenarios

Strikethrough

~ vs. ~~, inconsistent between models.

Skill session log - upgrading-stylelint, OpenCode Kimi K2.5, django-pattern-library

Pull request: django-pattern-library#275

Discussion: Feedback needed: agent skill for Stylelint upgrades #49

Package quadrants diagram

quadrantChart
    x-axis Low downloads --> High downloads
    y-axis Low stars --> High stars

    quadrant-1 "Popular"
    quadrant-2 "High visibility"
 quadrant-3 "Under the radar"

name	upgrade-wagtail
description	Plan and optionally execute an upgrade to a new Wagtail release, including all needed dependency updates.
license	MIT

Upgrade Wagtail

Usage:

Generated with OpenAI Codex (gpt-5.1-codex-max)

Repository Guidelines

Project Structure & Module Organization

django/: main framework packages; app directories mirror Django’s built-in contrib/apps layout.
tests/: Python test suite mirrors django/ paths (e.g., tests/forms_tests/ for django/forms/); runtests.py drives execution.
docs/: Sphinx sources; see docs/README for building HTML and make targets.

Role	Unacceptable	Capable	Adoptive	Transformative
Engineering	Calls AI coding assistants “too risky” Has never tested AI-generated code Relies only on Stack Overflow snippets	Uses ChatGPT / Copilot / etc for simple coding tasks (e.g. regex, unit-test stubs) Can explain how they prompt, review, and validate AI output	Chains LLM calls with fallback + retry logic Adds eval tests to flag hallucinations Knows Claude Code, Cursor, Windsurf, etc Can walk interviewers through prompt tweaks, token limits, code review	Ships LLM-powered features, monitors live metrics, and refines based on user feedback Builds an AI-first dev pipeline (guardrails, RAG docs, etc) that cuts down PR cycle time
Product	Dismisses AI as hype, showing no curiosity about user valuePRDs and prototypes lack any AI concepts or experiments	Uses ChatGPT to draft PRDs, story maps, and synthesize user-interview notesKnows basics (LLMs, embeddings, latency vs cost)

	#!/usr/bin/env -S uv run
	# /// script
	# dependencies = [
	# "python-frontmatter>=1.0",
	# ]
	# ///
	"""
	Generate a recap of issues assigned to specific milestones from the .issues/closed directory.

	This script parses Markdown files with YAML frontmatter and generates a formatted