Show HN: Git for LLMs – a context management interface (twigg.ai)

Twigg is a context-management interface for chat-based LLMs that turns linear conversations into a living tree of ideas. Instead of one long timeline, Twigg visualizes your entire project as an interactive tree where you can branch from any node, create parallel tangents, and move or delete nodes and branches to precisely control what context gets fed to the model. The goal is to prevent idea loss, reduce repeated context and token waste, and let multi-week projects live in a single navigable structure rather than scattered chats. For AI/ML practitioners and teams, this matters because it treats prompts and conversation state like version-controlled artifacts: you can explore alternate strategies without polluting the main thread, merge useful branches back in, and audit or reuse earlier contexts. Key technical implications include finer-grained prompt management (reducing irrelevant context and off-target outputs), lower token costs through selective context feeding, improved reproducibility of experiments and prompt engineering, and a collaboration model that scales beyond ephemeral chat threads. Twigg’s tree-based UI makes it easier to iterate, compare approaches, and maintain a clean, evolving knowledge graph for long-running AI projects.