Peter Steinberger Speeds Up Git Worktrees ~80% Using Filesystem Clones

Creating a git worktree normally means copying every tracked file into a new directory, which gets slow and disk-hungry once a repository is large or an agent needs many worktrees running in parallel. Steinberger's tool replaces that copy step with native filesystem clone operations, the same copy-on-write mechanism behind fast snapshots on APFS, Btrfs, XFS, and ReFS, so new files only take up space once they diverge from the original. The result is roughly an 80% reduction in the time it takes to stand up a new worktree, plus a corresponding cut in disk usage, with the implementation written in Rust and a possible port into Codex under discussion. Coding agents that isolate parallel sessions with worktrees pay this setup cost every time they spin up a new task, so a faster, cheaper worktree path translates directly into less wall-clock time and less wasted disk space for anyone running several agent sessions at once, particularly on filesystems that already support copy-on-write clones.