Jujutsu Primer
Jujutsu (jj), a Git-compatible version control system with a different data model and UX, is drawing interest for features like automatic snapshots, powerful rebasing, `jj undo/redo`, and treating the “staging area” as just another mutable commit. Commenters argue it can make complex workflows—stacked PRs, large monorepos, heavy rebasing, and AI-assisted coding—simpler and less error‑prone, though some find Git’s mature ecosystem, staging model, and tooling (especially LLM and forge integrations) sufficient or easier to live with. Adoption hurdles such as GitHub/GitLab centric workflows, missing features like Git LFS, and the need to relearn workflows are seen as the main reasons many experienced Git users hesitate to switch.
Overall sentiment
- Many see jj (Jujutsu) as a cleaner, more powerful UX over git, especially for complex histories and stacked work.
- Others feel git is “good enough,” deeply internalized, and jj’s benefits don’t justify switching or ecosystem friction.
Core model: jj vs git
- jj uses a DAG of mutable “changes” only; git has immutable commits + staging area + stash.
- Advocates say this unified model makes everything version-controlled (index, stash-equivalents, bookmark moves) and simplifies mental load.
jj undo/jj redo, the operation log, and automatic snapshots are highlighted as major quality-of-life improvements.- Conflicts don’t block operations; rebases keep going and conflicts can be fixed in follow-up commits, with remembered resolutions.
Workflows & staging
- Common pattern: treat the current change as a “staging commit,” then use
jj split/jj squash(often interactive) to carve work into logical commits. - Some describe stacked-PR / multi-branch workflows as dramatically easier: rebasing many PRs at once, keeping linear clean history, and jumping around without WIP commits.
- Critics prefer git’s explicit staging area, arguing it reinforces discipline and allows keeping debug harnesses and partial work out of commits; they see auto-tracking and no index as a regression.
- Others counter with
.gitignore,snapshot.auto-track, and squash workflows as sufficient to retain deliberate commits without the index.
Tooling, LLMs, and ecosystem
- Several use jj heavily with monorepos and custom backends; interest in jj-native forges not constrained by git formats.
- There is debate over git-format compatibility: some say it’s essential for adoption (Mercurial cautionary tale); others believe GitHub’s dominance has weakened and new backends are viable.
- LLM integration is mixed:
- Some report agents defaulting to git, making jj painful until they added skills, hooks, or AGENTS.md conventions.
- Others say recent models handle jj well, can manage deep commit stacks, conflict resolution, and squash workflows.
- UI tools (e.g., agent IDEs) often assume git worktrees and can misbehave with jj workspaces; support is improving but not “just works” yet.
- jjui and editor integrations are praised for making jj’s graph operations and splitting/squashing much more approachable.
Limitations & open issues
- Pain points: auto-bookmark naming, bookmark management, immature docs/UX in places, sharp edges with workspaces, lack of git-lfs support, and limited “critical mass” in tools.
- Some argue everything jj does could be scripted on top of git; they see a rewrite as unnecessary abstraction rather than a must-have.