Sources & methodology
The archive is built only from legally reusable open sources plus our own original analysis. Every record is traceable to where it came from.
Sources
| Source | What we use | License / terms |
|---|---|---|
| arXiv | Titles, authors, abstracts, identifiers, categories | Descriptive metadata under CC0 1.0. Full texts stay on arXiv; we link to them. Thank you to arXiv for use of its open access interoperability. |
| OpenAlex | Citation counts, venues, identifiers, open-access status | CC0 |
| Europe PMC | Biomedical metadata; abstracts only for articles licensed CC BY or CC0 | Per-article license, shown on each record |
| ClinicalTrials.gov | Trial registrations: phase, status, conditions, interventions, sponsors | U.S. National Library of Medicine data |
How dossiers are made
- Candidate selection. For each topic we rank the literature by influence (citations) and recency, combining several sources and deduplicating by DOI and arXiv id.
- Synthesis. AI research agents write an original state-of-the-field synthesis grounded in the indexed literature: results, benchmarks, timeline, open problems, glossary.
- Citation resolution. Every citation must resolve to a real record (arXiv id or DOI) in the archive. Unresolvable citations block publication.
- Adversarial verification. Independent verifier agents check each claim against its cited sources and flag unsupported or overstated statements, which are fixed or removed.
- Versioning. Dossiers carry a version, an update date and a verification record. They are refreshed as the field moves.
Dossiers are AI-assisted syntheses. They are checked carefully but can still contain errors. Treat them as an expert-curated map of the literature and confirm critical facts in the primary sources they cite.
Contact
Corrections, takedown requests and partnership enquiries: see security.txt for the contact address.