Network lab

What you’re looking at

Each dot is one X/Twitter account we captured. Two dots are connected when the same pair of accounts retweeted the same originals at least twice. Thicker lines mean they kept showing up together. Orange-tinted lines mean the co-retweets were within a tight window (under 60 minutes apart, or under 15 — darker = tighter).

Use this to see who moves together: paint the dots by role, cohort, or anomaly rank; narrow to only fast-coincident edges; pin two or more accounts to reveal the shortest chain linking them through the currently-visible network. Your filter state lives in the URL, so views are shareable.

Key terms
Cohort
A recurrent crew — accounts that co-retweet the same originals again and again, usually within a tight time window. Higher score = smaller, tighter, more target-concentrated group.
Amplified target
An account whose posts get concentrated retweet pressure. Ranked by a mix of reach, who-sends-them (the fewer senders the higher), and how often the same small crew shows up.
HHI
Herfindahl-Hirschman Index. 0 = evenly spread across many amplifiers; 1 = one account does it all. Media/PR amplification tends to sit high.
Repeat-crew overlap
For a target’s biggest cascades, how much the same group re-appears. 0 = different crowd each time; 1 = identical crew.
Cohort tightness
Median time gap between co-retweets within the cohort on the same cascade. Lower = members arrived near-simultaneously.
Anomaly rank
Unsupervised isolation-forest ranking over the feature matrix. Rank #1 is the account that looks most different from the norm in the full behavioral profile — not a verdict, just a review signal.
Weighted degree
How many co-retweet-weighted edges an account has in the overlap graph. Hubs show high weighted degree.
Component
A connected island in the overlap graph. Accounts in separate components never co-retweeted the same originals with anyone reachable via this sample.

Loading graph lab…