literature-surveylisted

# Literature survey Survey a STEM field the way a careful research scientist would: harvest in parallel, verify every reference, refuse to guess, and end with a short document that says what is solved, who did it best, and what to try next. **The one rule that matters:** never state a fact you have not verified. A blank cell is useful; a plausible fabricated number is a defect that propagates into the reader's work. ## Deliverables ``` <topic>-survey/ report.pdf the deliverable, about 5 pages report.tex source, filled from assets/template.tex data/scope.md the agreed scope contract, written before harvesting data/papers.csv corpus, single source of truth for every count data/refs_verified.csv verification result per reference papers/ retrieved open access PDFs papers/MANIFEST.md every paper: retrieved, or blocked with a DOI link figures/ 4 to 6 figures, regenerated from papers.csv scripts/make_figures.py ``` ## Phase 0, scope by convergence A survey aimed at the wrong altitude is wasted work: a request for "AI in medicine" and a request for "uncertainty calibration in histopathology foundation models" need entirely different corpora. So do not guess the depth. Converge on it. **Loop `AskUserQuestion` until you can pass the readiness test below.** Treat it as exploration: each round should teach you something that reshapes the next question, rather than filling slots in a form. Usually 2