Juju-chu: Guide to Jujutsu version control for AI coding agents

Yuka Ooka brings twenty years of Git-based development and authorial success (Riakuto! exceeded 40,000 copies sold) to a problem that only recently became acute: the mismatch between how Git was designed and how AI coding agents work. Where Git assumes human-paced, deliberate commits, agents generate, rewrite, and discard large volumes of code in minutes. Ooka's own workflow shifted when she began using AI agents daily; staging became a bottleneck, commit archaeology through reflog became routine, and history rewriting—already risky—felt paralyzing when an agent had just rewritten major sections. Jujutsu's design answers this directly: automatic snapshots mean nothing is ever lost, the operation log replaces reflog archaeology, and rewriting history stops being something to brace for.

The book's pedagogical choice—dialogue between a senior engineer and a junior who asks the questions readers form in their heads—draws on proven teaching method (Socrates, The Little Schemer) and on Ooka's own success with Riakuto!. It targets not Git masters but working developers who feel friction: those whose AI agents have caused them to lose uncommitted work, those who dread rebasing, and those who spend hours splitting sprawling agent output into clean, reviewable changes. Critically, adoption is low-risk: Jujutsu sits on top of Git, so a developer can switch alone, push ordinary commits to GitHub, and let teammates remain on Git. That reversibility—and the fact that the decision is purely individual—removes institutional barriers to trying the tool.