TLDR
A finite state machine manages 53 deduplicated leads drawn from three sources: 20 from cross-domain synthesis, 33 from anomaly-driven convergence, and 2 manually promoted entities -- deduplicated to 53 (PAPER TRAIL Project, 2026a). Each lead carries Monte Carlo robustness scores where available and expires after 14 days without advancement, preventing stale leads from cluttering the queue.
What a Lead Queue Does
An investigative pipeline that produces leads without managing them will drown in its own output. The lead queue is a structured system for tracking which entities warrant further investigation, where those leads came from, how confident the system is in each one, and whether they have been acted upon (PAPER TRAIL Project, 2026a).
The queue implements a finite state machine (FSM) with five states: NEW (initial intake), INVESTIGATING (analyst has begun work), CONFIRMED (evidence upgraded to finding threshold), DOWNGRADED (evidence insufficient or lead identified as artifact), and STALE (14 days without state transition). Every lead enters as NEW and must be explicitly moved to another state. The 14-day stale timer prevents leads from sitting indefinitely in NEW without review (PAPER TRAIL Project, 2026a).
Three Sources
Synthesis leads (20). These are entities from the cross-domain synthesis engine (Script 25b) that appear across multiple analytical domains with convergent evidence (PAPER TRAIL Project, 2026b). They include the nine cross-domain leads -- Jeffrey Epstein, Ghislaine Maxwell, Southern Financial LLC, Bella Klein, Financial Trust Company, Darren Indyke, Butterfly Trust, JEGE Inc, and NYSG LLC -- plus 11 additional entities with cross-domain profiles that met the lead threshold. Each synthesis lead has Monte Carlo robustness scores attached: a probability of being classified as a finding under parameter uncertainty, and 90% confidence intervals on the adjusted confidence score.
Convergence leads (31). These are entities identified by Script 30's anomaly-driven lead discovery, which runs five independent detection modules -- hidden brokers, wire timing anomalies, orphan clusters, community anomalies, and document dark matter -- and flags entities that appear in two or more modules (PAPER TRAIL Project, 2026c). All 33 entities with multi-module convergence were loaded into the lead queue, with deduplication against synthesis leads reducing the total from 55 (20+33+2) to 53. The top convergence lead is UBS AG, the only entity appearing in three detection modules -- document dark matter, orphan clusters, and wire timing (PAPER TRAIL Project, 2026c).
Manual leads (2). Joe Pagano and Kevin Law were manually promoted based on profile analysis that identified investigative significance beyond what automated modules detected (PAPER TRAIL Project, 2026c).
Joe Pagano is classified as HIGH priority: an inner-circle financial associate with 1,938 corpus mentions, documented island access (EFTA00374252), presence in both unredacted and redacted flight logs, and co-occurrence with Leon Black (95 times), Woody Allen (71), Boris Nikolic (57), and Reid Weingarten (35) (PAPER TRAIL Project, 2026d). His association with "Curve Capital" and a role as NYU admissions intermediary appeared in the anomaly profiles.
Kevin Law is classified as MODERATE priority: a USVI and EB-5 business contact with 1,835 corpus mentions, post-conviction contact (December 2015), and co-occurrence with Joe Meli (54 times -- Meli was convicted in a $102 million Ponzi scheme), Jes Staley (47), and USVI references (95) (PAPER TRAIL Project, 2026d).
Deduplication
Leads from different sources frequently identify the same entity. The queue deduplicates by normalized entity name, merging source metadata when the same entity appears in both synthesis and convergence outputs (PAPER TRAIL Project, 2026a). Joe Pagano, for example, appeared in both the convergence module (hidden brokers and orphan clusters) and was manually promoted -- his lead queue entry carries the source label "convergence+manual" and aggregates evidence from all sources.
After deduplication, the 53 leads represent unique entities drawn from 20 synthesis leads, 33 convergence leads, and 2 manual leads, with overlapping entities merged and their source metadata combined (PAPER TRAIL Project, 2026a).
Monte Carlo Attachment
The 20 synthesis leads carry Monte Carlo robustness scores from the 5,000-iteration sensitivity analysis (PAPER TRAIL Project, 2026e). Each score represents the probability that the lead would be reclassified as a finding under parameter uncertainty. Across all 20, this probability is 0.0 -- no parameter draw elevated any lead to finding status. The 90% confidence intervals on adjusted confidence range from 0.042 to 0.296, all below the 0.75 finding threshold.
The 33 convergence leads do not carry Monte Carlo scores because they were generated by a different analytical pathway (network topology and anomaly detection rather than evidence chain evaluation). Attaching robustness scores to convergence leads is a planned enhancement.
Current State
All 53 leads are in NEW state as of the queue's creation date (PAPER TRAIL Project, 2026a). The stale timer gives each lead until March 25, 2026 before automatic transition to STALE. The queue is a snapshot of the system's current investigative priorities, not a static document -- it is designed to be worked, with leads advancing through states as analysts evaluate them.
References
PAPER TRAIL Project. (2026a). Lead queue state machine [Data set]. _exports/lead_queue/lead_queue.csv, lead_queue_summary.json
PAPER TRAIL Project. (2026b). Cross-domain synthesis leads [Data set]. _exports/synthesis/leads_queue.csv
PAPER TRAIL Project. (2026c). Anomaly-driven convergence leads [Data set]. _exports/anomaly_leads/convergence_leads.csv, anomaly_summary.json
PAPER TRAIL Project. (2026d). Lead profiles [Data set]. _exports/anomaly_leads/LEAD_PROFILES.md
PAPER TRAIL Project. (2026e). Monte Carlo lead robustness results [Data set]. _exports/synthesis/mc_lead_robustness.csv
This investigation is part of the SubThesis accountability journalism network.