I feel like we’re over-capitalizing a problem that could be solved with better protocols. If the "Git successor" is just a wrapper to help agents not hallucinate their own worktrees, it feels like a very expensive solution to a context-window management problem.