user-in-search-of-a-name gives Mike a matter-long memory
A fork that turns the assistant from a per-chat helper into a system that remembers a matter across every conversation.
Most legal-AI assistants forget everything the moment a chat ends. This fork builds a "Matter OS" layer that fixes that: a curated knowledge base of structured entries - facts, parties, key dates, clauses, positions, playbooks, sources - that lives at the matter level. Entries can sit in a personal library and be cloned into specific matters, and the assistant can read and search them while it works, with the most relevant ones fed straight into its context.
The smart restraint: when the assistant spots something worth saving, it suggests the entry for a human to approve rather than writing to the record itself. There's also a unified activity feed pulling chats, document versions, edits and reviews into one timeline. One caveat worth knowing before you lean on it: the search is fairly basic today, so as a matter's knowledge grows it can miss older entries.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?