Literature Hypothesis Extractor
Turn a paper into distinct, falsifiable, provenance-tagged claims the platform can test.
How to read these claims. Every claim here is a formalization written by AI reasoning (provenance tier inferred) — a faithful restatement of what the paper says, not a new measurement. The paper’s own basis for each claim (experimental data, computation, or its own hypothesis) is shown separately. Statements too vague to make falsifiable are recorded as abstentions, never invented.
Evidence Harvester — the data trail behind each paper
PubMed tells us what scientists said. This station follows each mined paper back to what was actually measured — the 3D structures, sequencing, proteomics, datasets and code it deposited — then links and honestly classifies each one. It runs automatically on every newly mined paper. Nothing is fabricated: a structure is marked verified only when it actually resolves live on RCSB, and raw-omics reanalysis is honestly parked (no pipeline exists yet).
Try (Bose et al. 2013, activating HER2 mutations).