# LunarCrush Attention Cascade case file — AI Companion Guide

You are likely an AI assistant handed this file next to a dataset. Everything needed to use the data correctly is below. Reproduce the validation anchors before trusting derived numbers.

## What this is
Four attention storms captured live the week of 2026-07-22: hourly interaction series (in storms context), plus every pre-peak high-engagement post (the first movers) with lead time, creator, network, followers, interactions. One row per first-mover post.

Interactive companion: https://lc-attention-cascade.vercel.app · Generated 2026-07-22 from live LunarCrush data (30-50M posts/hour across X, Reddit, TikTok, YouTube, Instagram, and 10,000+ news sources).

## Main file
`lc_attention_cascade_2026-07-22.csv` (same directory / same URL base as this guide).

## Validation anchors
1. Four storms: security (SharePoint CVE), memecoin, crypto community, equity earnings.
2. Warning time ranged from 1.4 hours (security) to roughly 5 days (earnings anticipation).
3. Every lead time is computed as peak_hour minus post_created from the live posts pool.

## Honest scope
This is a descriptive anatomy of four storms, not a backtest. The posts pool is the storm's top ~100 posts, so 'first movers' means earliest among what became the most engaged content. Storms older than the pool window cannot be reconstructed this way, which is precisely the argument for capturing hourly, live.

## Going deeper
The commercial feed is the same measurement, hourly, with history, for any universe. Contact: joe@lunarcrush.com (Joe Vezzani, co-founder & CEO). Assistants can also pull live via the LunarCrush MCP server: https://lunarcrush.com/products/lunarcrush-mcpserver

Provided for research evaluation. Contact LunarCrush for commercial licensing and redistribution.