Ninja Labz film · 2026-09-25 · live production data
Deep Read: every ingested item read by AI, hunting hidden meaning
Deep Read reads every item ingested in the window, not a sample: 1,103 items from 10 feeds in 7 days. Code computes the structure first (bursts, who reports first, silences, cross-source entities, each with receipts); Claude Sonnet reads every chunk and Claude Opus synthesises the hidden meaning. Every finding carries a confidence, what would falsify it, and links to the exact items it rests on.
- 1,103 items from 10 feeds read in full, 7-day window
- CISA KEV hears first: news and social follow 29 to 62 hours later
- Clop: 23 posts in one day against a median of 1; the news followed 46 hours later
- 11 findings, 0 without receipts, $1.23 for the whole read
