Claude Code setup leans on CLAUDE.md

The interesting part here is not the install path, it’s the memory model: Claude Code becomes much more useful once the repo itself teaches the agent how to behave. That makes this feel less like autocomplete and more like a persistent project worker.

• –`CLAUDE.md` is the difference between a one-off session and usable project memory
• –The video emphasizes working inside an actual VS Code folder, which makes the workflow feel practical rather than demo-only
• –Reading files, writing outputs, and keeping instructions in-repo are the core behaviors that matter for daily use
• –This is a strong onboarding/tutorial signal for teams that want repeatable agent behavior without re-explaining context every session
• –The framing suggests Claude Code’s real value is structured execution, not chatty brainstorming