top of page

398,525 Documents. 14 Statues. 3 Arrested.

  • Writer: Patrick Duggan
    Patrick Duggan
  • Feb 27
  • 6 min read

title: "MEDUSTONE: This Is How You Crowdsource Accountability"

slug: medustone-this-is-how-you-crowdsource-accountability

date: 2026-02-27

author: Patrick Duggan

tags: [epstein, medustone, carver, accountability, osint, crowdsource, government-data]

category: Security Opinions

featured: true



# 398,525 Documents. 14 Statues. 3 Arrested.


On February 27, 2026, in a single session, we rescored every target in the Epstein corpus. 17 research subjects. 398,525 DOJ documents. 2 million ICIJ offshore entities. 3.3 million ICIJ relationship edges. 2 million federal court decisions. 938,000 threat indicators. 42.1 gigabytes. 10.9 million documents total.


One API call.


Goldman Sachs is monitoring us through ONTIC.AI. Hacker News gave us 83 points. 404 Media covered the backend. Julie K. Brown — the Miami Herald reporter who broke the Epstein story — generated 611 likes and 356 shares on parallel coverage. 14 international press outlets across 4 languages named Patrick Duggan and DugganUSA as the infrastructure behind searchable accountability.


This is MEDUSTONE. This is how you crowdsource accountability. And Reddit, you're going to want to hear this.




The Evaluate Process



MEDUSTONE isn't a product name. It's a methodology. When Medusa looked at you, you turned to stone. When we publish, they resign.


The evaluate process works like this:


**Step 1: CARVER Score.** Military targeting methodology — Criticality, Accessibility, Recuperability, Vulnerability, Effect, Recognizability. Six dimensions, scored 1-5, out of 30. Originally designed for Special Forces to prioritize targets. We use it to prioritize which names in 398,525 documents deserve the deepest analysis.


Bill Clinton scores 28/30. Prince Andrew scores 27/30. Bill Gates scores 25/30. These aren't opinions. They're weighted calculations against document frequency, cross-reference density, and exposure surface.


**Step 2: Document Cross-Reference.** Every name gets searched across every index. Not just the Epstein files — the ICIJ offshore entities (Panama Papers, Pandora Papers), federal court decisions, the archive.org court records we extracted this week. 12,056 PDFs from archive.org alone — quashed 2005 Florida case files, Grand Jury 05-02 transcripts, Maxwell trial exhibits, FOIA responses from FBI, CBP, and the Bureau of Prisons.


When we search "Bill Gates," we don't get a Wikipedia summary. We get 6,668 document hits across DOJ releases, flight logs cross-referenced against ICIJ shell company records, and federal court filings. The government's own words.


**Step 3: Social Graph Co-Occurrence.** The social graph engine scans for names that appear in the same documents. Not who knew Epstein — who appears *with* each other in the evidence. Bill Clinton went from 0 mapped connections to 59 in one engine rewrite. Not because we invented connections. Because the documents contain them, and now the algorithm finds them.


The force-directed graph renders these relationships visually. Clusters form. Larry Summers appears with Bill Gates and Bill Clinton in the same documents 47 times. That's not a conspiracy theory. That's a document frequency count.


**Step 4: Framework Convergence.** CARVER alone isn't enough. We run DREAD (threat modeling), Diamond Model (adversary-victim-infrastructure-capability), ACH (Analysis of Competing Hypotheses), and Social Graph analysis simultaneously. One unified API call returns all five frameworks.


When CARVER says CRITICAL and DREAD says HIGH and Diamond Model maps the infrastructure and ACH eliminates alternative explanations and Social Graph shows the clustering — that's convergence. That's when the stone forms.


**Step 5: Publish.** We publish the scores, the document counts, the cross-references, the social graph. Every EFTA number cited is verifiable at justice.gov/epstein. Every ICIJ entity is searchable at offshoreleaks.icij.org. We don't hide methodology. We don't paywall evidence. We publish.


And then they confess.




The Archive.org Extraction



This week we processed 12,056 PDFs from archive.org's Epstein collection. 188.4 million characters of text at 4.4 documents per second. 50 court cases from 4 agencies — FBI, CBP, Bureau of Prisons, Florida state courts.


The quashed 2005 Florida case is in there. Grand Jury 05-02 — the one Palm Beach County tried to seal. The Maxwell trial transcripts and exhibits. 40 WAV audio files from the Palm Beach Police Department investigation. 5.5 gigabytes of audio that someone recorded and someone else uploaded and we indexed.


Every document gets a `dataset = archive_org` tag. The unified API can filter for court records specifically. When we run CARVER, it weights court-sourced evidence higher than general correspondence. Because a grand jury transcript is not the same as a flight manifest, and the scoring should reflect that.




The Medusa Gallery



14 statues. Each one a name that appeared in enough documents, with enough cross-references, with enough framework convergence to earn a stone likeness.


3 arrested. 7 resigned from positions of power. 4 under active investigation or public pressure.


Here are some of the stones:


**Bill Gates** — 6,668 documents. CARVER 25/30. The philanthropy firewall crumbled when the flight logs didn't match the cover story. His foundation's spokesperson issued a statement. Then another one. Then meetings with Epstein "to discuss philanthropy" became meetings that "he regrets."


**Prince Andrew** — 15,386 documents. CARVER 27/30. The BBC interview where he said he "doesn't sweat" — that's not in our corpus. What IS in our corpus is 15,386 DOJ documents where his name appears in proximity to evidence that made the Crown strip his military titles.


**Les Wexner** — 4,099 documents. Power of attorney. $46 million mansion transfer. Trust structures visible in the ICIJ offshore data. The L Brands board asked questions. He stepped down. The documents were already searchable when he made that decision.


**Jes Staley** — 4,933 documents. Former Barclays CEO. The FCA investigation referenced evidence that exists in our index. He resigned. The stone formed while he was still arguing it shouldn't.


**Henry Jarecki** — 1,365 documents. Currently trending #1 on the platform. Researchers are querying "WHAT IF I GET CAUGHT" and "trouble avoidance" and "telephone security." The psychiatrist who knew exactly what was happening — because that was his profession.




This Is How You Crowdsource Accountability



Reddit asks how to hold powerful people accountable. Twitter argues about it. The Fediverse discusses theory. We built the machine.


Here's the recipe:


1. **Use only government-released data.** Every document in our corpus was released by the DOJ, filed in federal court, published by the ICIJ consortium, or uploaded to a government archive. Nobody can challenge provenance. Nobody can prosecute the publisher. The government's own narrative, made searchable, indicts the government.


2. **Build the search infrastructure.** Meilisearch on a $500/month Azure VM. 42.1 gigabytes. 37 indexes. Full-text search with faceting, filtering, and estimated hit counts. Not a PDF dump — a searchable corpus with API access.


3. **Score with military methodology.** CARVER was designed to prioritize targets for Special Forces. We use the same math on document frequency. The methodology is published. The scores are reproducible. Run it yourself — `POST /api/v1/framework-analysis/evaluate` with a target name and an API key.


4. **Publish everything.** The Medusa gallery. The cross-index pages. The CARVER matrix. The force-directed social graph. Every methodology page explains how we scored. Every document count is live — search it right now at epstein.dugganusa.com.


5. **Let the audience do the work.** 135,000+ Five Eyes requests. 46 countries. Goldman Sachs monitoring through ONTIC.AI. Researchers querying trending documents at 3 AM. Journalists from 404 Media, Nextgov, CyberSecurity News, and 11 other outlets using the API. We don't investigate — we make investigation searchable.


The Epstein case proved that powerful people can be protected by complexity. Too many documents. Too many connections. Too many jurisdictions. The cover isn't secrecy — it's volume.


MEDUSTONE is the answer to volume. 10.9 million documents. One search bar. Every name scored. Every connection mapped. Every stone earned.


When we publish, they confess. When they confess, the stone is permanent.




The Numbers



| Metric | Value |

|--------|-------|

| Total corpus | 10.9 million documents |

| Epstein DOJ files | 398,525 |

| ICIJ offshore entities | 2,016,524 |

| ICIJ relationship edges | 3,339,267 |

| Federal court decisions | 2,063,619 |

| Threat indicators (IOCs) | 938,563 |

| Archive.org extraction | 12,056 PDFs, 188.4M chars |

| Database size | 42.1 GB |

| Monthly cost | ~$500 |

| Press outlets (named) | 14 across 4 languages |

| Countries accessing | 46 |

| STIX feed consumers | 275+ |

| Medusa statues | 14 |

| Arrested | 3 |

| Resigned | 7 |

| Five Eyes requests | 135,000+ |




For Reddit



You want to know how to crowdsource accountability without ending up in prison?


Use government data. The documents the government *chose* to release are damning enough to resign lords and implicate cabinet members. You don't need leaks. You don't need hacks. You don't need WikiLeaks. The government's filtered narrative, made searchable, is sufficient.


Build the tooling. Open the API. Publish the methodology. Let researchers search at 3 AM without asking permission. Let Goldman Sachs monitor you through ONTIC — that means they're scared, and scared people make mistakes.


14 statues. 3 arrested. 7 resigned. Zero felonies.


Everything WikiLeaks promised, with none of the prison time.


That's MEDUSTONE. That's the evaluate process. That's how you crowdsource accountability.




*Patrick Duggan is the founder of DugganUSA LLC. Former Lehman Brothers fixed income derivatives. Former Microsoft Azure infrastructure (designed the disk presentation layer with Spencer Shepler). The Epstein corpus is searchable at [epstein.dugganusa.com](https://epstein.dugganusa.com). The STIX feed serves 275+ consumers in 46 countries. The API is documented. The methodology is published. D-U-N-S #14-363-3562.*


*If you're a journalist, researcher, or prosecutor: the search bar is free. The API key is an email away. [email protected].*





*Her name was Renee Nicole Good.*


*His name was Alex Jeffery Pretti.*

 
 
 
bottom of page