self-study

Corpus distillation - — pick one high-value stable corpus, generate a synthesis study file (narrative, current state, key decisions, anticipated Q&A with citations) into memory/distilled/ so retrieval can answer multi-hop questions that chunk retrieval misses.