Episode 6: Computational Forensics at Scale
6 parts
Vision models reading handwritten flight logs, network topology across 2.38 million entities, and changepoint detection on decades of documents. Each part builds on the last — investigative analysis drawn from court filings, financial records, and released documents. Browse all episodes or the full investigation archive.
- Part 1 4,286 Flights: How a Vision Language Model Read Handwritten Flight Logs Network & Temporal
Traditional text extraction failed completely on Epstein's handwritten flight logs. A vision language model -- a type of AI that interprets entire page...
- Part 2 125,620 Communities: Network Topology of 2.38 Million Entities Network & Temporal
A grouping algorithm that finds clusters of closely connected entities -- called Leiden community detection -- applied to 29.5 million entity relationsh...
- Part 3 889 Breakpoints: Detecting Temporal Shifts in Document Activity Network & Temporal
An algorithm that finds sudden shifts in document activity patterns -- called PELT -- detected 889 verified breakpoints in document activity time series...
- Part 4 Information Surprisal: Finding the Documents Nobody Searched For Network & Temporal
A scoring method that flags documents containing unusual combinations of names -- combining two established information-theory measures called IDF and P...
- Part 5 610 Breakpoints: How PELT Recalibration Tightened Temporal Analysis Network & Temporal
A penalty sweep across 20 logarithmic steps reduced the corpus change-point count from 889 to 610, with the optimal penalty of 0.2069 selected by elbow...
- Part 6 Flight Domain Integration: 13 Passengers Who Also Appear in Financial Records Network & Temporal
After a vision language model extracted 4,286 flights and 392 passenger names from handwritten logs, name deduplication and cross-domain matching identi...