A primary source is close to the event or subject being studied: a protocol specification, transaction record, court filing, original announcement, or contemporaneous message. The Library of Congress emphasizes original materials and careful interpretation. In crypto research, the appropriate source depends on the claim. A whitepaper can establish what an author proposed; deployed code can show implemented rules; neither alone proves customer adoption or a company's financial condition.

Suppose an article says a regulator approved a company because its lawsuit was dismissed. Read the actual announcement and identify its stated reason. The SEC's Coinbase dismissal announcement explicitly separates its policy decision from an assessment of the merits. That distinction changes the conclusion. Record the document date, author, procedural status, and any limitations, then write the narrow claim the record supports rather than preserving the more dramatic wording of a headline.

A project's own page is primary evidence of its claims but not independent confirmation that those claims are true. Marketing, selective disclosure, preliminary research, and outdated documentation can all be primary material. Compare sources with different roles and inspect whether they genuinely support the same proposition. For a reproducible note, save the exact URL and retrieval date, identify the relevant passage or section, and distinguish direct observation, attributed statements, inference, and unresolved uncertainty.