How ThreadEast works
ThreadEast is a discovery and context layer over selected public trend pages, not an archive or a truth score.
Minimal collection
We collect list-page metadata: title, rank, public URL, source heat label, and capture time. We do not store article bodies, comments, videos, account data, or user profiles.
Direct provenance
Every signal names its source and links to the original public page. Source rankings describe attention on that platform; they do not prove that a circulating claim is true.
Bounded AI
Automation translates only known titles and classifies broad categories. It must not invent people, dates, causes, outcomes, or background. Model and prompt versions are retained internally.
Editorial review
Sensitive subjects and low-confidence output enter a human review queue. Editor-reviewed explainers are clearly marked; machine-translated single signals remain distinct.
Responsible sources
Sources must be public, policy-checked, and registered before collection. We do not automate logins, solve CAPTCHAs, evade access controls, or collect from sources that prohibit it.
Corrections and removal
We correct material errors visibly and retain correction metadata. Rights holders and source operators can request source-specific review or removal.
View correction process