literature-trackinglisted

# Literature tracking Three literature firehoses, one deduplicated stream, ranked against what the user actually works on. `scripts/track.py` does the deterministic half — query, normalise, deduplicate. The judgement half is **not** its job: deciding what is worth the user's attention, and saying why. Do not try to script that. A model reading titles and abstracts beats a keyword filter at it, and can explain its reasoning afterwards. ## Workflow ### 1. Establish the research profile Ranking is impossible without topics. Take them from the conversation, from project memory, or ask. A usable profile has: - **topics** — what they study, in their words ("cryo-EM of membrane transporters", not "biology") - **methods or systems** — techniques and target molecules that matter - **exclusions** — the neighbouring work they keep getting served and do not want If they have named seed papers, treat those as the strongest signal available. ### 2. Choose the window and sources Default to the last 7 days across arXiv, bioRxiv, PubMed and Europe PMC — that is exactly what `--sources` defaults to. Widen to `30d` if the topic is slow-moving or the last check was long ago. **medRxiv is off by default.** Clinical questions need `--sources arxiv biorxiv medrxiv pubmed europepmc` *as well as* `--medrxiv-categories`; passing the categories alone is a usage error and the script stops rather than quietly searching nothing. ### 3. Map the profile onto each source — they do not accept the