flagship storyEchoTime

AI attention breakout analogs

A showcase story for asking which historical attention breakout DeepSeek looked most like over its first breakout window.

Open assetsReport + story + card
Use caseExplainable similarity demo

Why this demo travels

Series overlay
EchoTime series previewSeries previewDeepSeekThreads

The closest analog stays shape-aligned enough that the story is visible before you read the verdict.

Similarity radar
Similarity radarSimilarity radarRadar over the time-series metrics. Read it together with Pearson, Spearman, and mutual info.ShapeDTWTrendDerivativeSpectral

The radar shows which time-series metrics keep supporting the analog and which ones stay weaker.

Rolling component mean
Rolling component meanRolling component meanmean=0.51, min=0.20, max=0.93 across per-window metric means

Windowed component means show whether the analog survives beyond the first surge of attention.

Open the assets

Open the HTML report Open the social card Open the Python source Open the local CSV snapshot Open the upstream public source

Data note

This story uses a frozen local CSV snapshot plus the upstream Wikimedia pageviews endpoint so the result stays reproducible.