What's this about?
swyx went viral complaining that git worktrees — a common way to give each coding agent its own isolated copy of a repo — duplicate the `node_modules` dependency folder every time, until you're carrying 20GB of the exact same files.
What exactly is a "worktree"?
A git worktree is a second (or third, or tenth) working copy of the same repository, so multiple tasks — or multiple AI agents — can each edit code in isolation without stepping on each other. The catch: each copy re-downloads the project's dependencies (`node_modules`), which for JavaScript projects can be enormous and identical across copies.
Why it matters
Multi-agent coding needs isolation, but naive isolation wastes huge amounts of disk. It's an early, concrete scaling pain of the "many agents working in parallel" future everyone's building toward.