1 séries · 12 articles
Outils, patterns et workflows pour expédier du code prod avec des partenaires IA — de l'autocomplétion aux agents autonomes.
Where this is going in 2026 and beyond. What I'd bet on, what I would not, and the line where I get sceptical of the hype.
Once you've outgrown the defaults: hooks for deterministic side effects, MCP servers for org-specific data, custom tools, and system-prompt surgery.
How engineering teams are actually integrating Claude Code today. The shared .claude/ folder, the review rituals, and the anti-patterns I keep seeing in the wild.
Letting Claude Code own the entire test loop. Including the parts that make engineers nervous: regressions, flakies, integration tests, and the stack-trace whisperer.
From Linear ticket to merged PR with Claude Code. A real, honest walk-through — what the prompt looked like, what the agent got right, what I caught in review.
Chaining sub-agents, running them in parallel, and the patterns for 'review-while-coding' without losing your mind. Where Claude Code starts to feel like a small engineering org.