Show HN: Paca – Lightweight Jira alternative for human-AI collaboration

That landed because a lot of people are already improvising this workflow. Several described running multiple git worktrees in parallel, one branch per task or pull request, then pointing Claude, Cursor, Zed, or other agents at each one while they supervise and review. Others said GitHub Issues, GitHub Projects, repo-local markdown files, Telegram topics, or custom internal tools are doing the job today, but barely. The pain is not "we need another kanban board." It is that agent work creates more parallel tasks, more state to track, and more ambiguity about whether output is actually finished or just plausible. The strongest thread running through the comments was that existing tools are mismatched to this new operating mode. Jira still has fans, mostly for its flexibility and richer workflow features, but many saw seat-based enterprise tools as too slow, too heavy, or too expensive for agent-driven development. GitHub works for many teams, yet it lacks higher-level coordination features people now want, especially project-level memory and better handling of non-code tasks across repos. Paca’s project chat drew interest for exactly that reason. There was also a clear warning under the enthusiasm. People increasingly customize these systems around their own habits, and LLMs make it cheap to build hyper-personal internal tools. That makes adoption harder for any shared product, because everyone’s "simple" setup is different. A Joel Spolsky quote about the 80/20 myth got traction as the right frame. Everyone only uses part of Jira, but not the same part. The practical opening for tools like Paca is not beating Jira at breadth. It is staying lean while exposing enough extension points, like the WASM plugin architecture, to let teams shape the missing 20 percent themselves.