Also want to acknowledge this approach has some limitations. Human judgement in sharing and managing context is still crucial. What I think memsprout does is provide transparency in shared context management. But the last mile is still (and should still imo) be human owned.

And I thought of a concrete example as a use case: acronyms, code names, etc. Every company I have been at has tons of these. In my first attempt of a repo based approach we used an .md file as a “glossary” of company terminology. But updating that file became a tedious, focus shifting, task. With memsprout as soon as I explain to claude that “StepLadder” is our new product line, claude can store that as a memory and immediately make that new insight available to the rest of my team’s agents.