What a day for Context Compaction! > Morph trained a dedicated model for Context Compaction and yes, this is one of many usecases where SLMs fit and they shd win. > Cursor RLmaxxed again to self-summarize Composer and it effectively reduces error by 50% in Context Compaction.