Sitemap
Pages
Articles (32)
- How to read a token-compression benchmark
- A $214 AI coding agent rewrite
- Doctolib's cross-verified AI adoption data
- FinOps applied to tokens: who owns the AI bill
- The trillion-dollar backdrop: what markets see that teams don't
- Measuring your own AI usage: three meters, three different answers
- The price-per-token lie: why cheaper models don't mean cheaper bills
- AI coding productivity at scale: 15-20% measured, 20x in one team
- You cannot depend on a single AI vendor
- Why AI ROI stays invisible, even to the people measuring it
- 2/2 · Claude selected my output style. Then ignored it
- 1/2 · Why I combined three Claude Code skills
- Campus IA in Fouju: the project in brief
- Campus IA à Fouju : le projet, en bref
- Campus IA in Fouju: the project in detail
- Campus IA à Fouju : le projet détaillé
- Portable agent configuration is a release system, not a shared folder
- Your CLAUDE.md is too long
- 6/6 · Portability becomes a Scale concern
- 5/6 · The AI instruction system is a product, not a config file
- 4/6 · The responsibilities around context engineering
- 3/6 · Mapping the token-reduction toolbox
- 2/6 · Diagnose and repair context drift
- 1/6 · Opposite AI results: what context can explain
- Claude is my second contributor: what real Git stats show
- Claude Code under the hood
- Choose your AI-assisted engineering starting path
- From afterthought to infrastructure: how AI config evolves in a real project
- AI velocity is bidirectional
- Non-technical to production in 10 days with Cursor AI
- UVAL: the protocol I built to stop accepting code I don't understand
- From VP engineering to solo builder: my honest take
Notes (2)
Guides (14)
- Moving the memory observer to a local model: five constraints nobody documents
- Jev : comprendre, comparer et tester le modèle de TypeSafe
- Jev: understand, compare and test TypeSafe's model
- Une seule source configure Claude Code et Codex
- One source configures Claude Code and Codex
- Back Market a sorti 245 développeurs d'Anthropic en direct
- Back Market moved 245 developers off direct Anthropic access
- Persistent memory: the six failures that never raise an error
- Context engineering: the L0-to-L5 playbook
- MCP servers: what they actually cost and when to use them
- Plan-execute: running 20+ agents on a migration
- Claude Code security: the attack surface nobody audits
- Claude Code setup, level by level
- TDD with Claude Code