One Guy, a Search Engine, and the Network the Government Tried to Bury
- Patrick Duggan
- Feb 25
- 9 min read
# One Guy, a Search Engine, and the Network the Government Tried to Bury
**Published:** February 26, 2026
**Author:** Patrick Duggan
*How a cybersecurity company in Minnesota applied military targeting methodology to 364,000+ Epstein documents — predicted what Congress would find before Congress found it — and watched six people turn to stone in 26 days.*
The Problem Nobody Wanted to Solve
In January 2026, the Department of Justice released the Epstein files. Over 136,000 documents — court filings, FBI reports, financial records, victim statements. They released them as scanned images. Unsearchable PDFs. No index. No database. No way to cross-reference a name in document 4,000 with the same name in document 97,000.
If you wanted to find out whether Steve Bannon was in the files, you could start reading. At one document per minute, eight hours a day, it would take you 284 working days. A full year, no weekends, no holidays, just reading.
This is not an accident. This is how you release information without releasing information. You put 136,000 documents on a website, hold a press conference, and let the sheer volume of the disclosure become its own redaction. Nobody can search what nobody can read.
We built the search engine.
$500 a Month
DugganUSA LLC is a cybersecurity company in Minnesota. Two people. We build threat intelligence infrastructure — the kind of systems that hunt malware, track ransomware gangs, and map criminal networks. The same tooling that tracks a Russian C2 server through six proxies can track a financial predator through six shell companies.
So we pointed it at the Epstein files.
One Azure virtual machine. One Meilisearch instance. One ingestion pipeline that converts scanned images into searchable text, extracts names and dates, and indexes everything into a full-text search engine.
Cost: $500 a month.
By February 2026, we had 364,000+ documents indexed — the DOJ release, the House Oversight Committee's Congressional investigation files, and the ICIJ's offshore entity database. Every document searchable. Every name cross-referenced. Every connection mapped.
But indexing isn't analysis. Search is a tool, not an answer. The question wasn't what's in the files. The question was what's missing from them.
The Switchboard
To understand the cipher, you need to understand what Epstein was.
He wasn't a billionaire. His actual wealth was never established — estimates range from $500 million to maybe zero, depending on which financial structure you unwind. He wasn't a hedge fund manager. His one known client was Les Wexner, and even that relationship defied normal financial logic.
What Epstein was, demonstrably, based on 139,619 government-released documents: a switchboard.
The same man who hosted Steve Bannon at 6:45pm on March 22, 2018, had hosted Ehud Barak at 6:00pm the same evening. Two days later, the schedule shows Joi Ito, Reid Hoffman, Woody Allen, Soon-Yi Previn, and Kathy Ruemmler — Obama's former White House counsel.
The same inbox that received Boris Nikolic's "mission accomplished for some" — Nikolic being Bill Gates' science advisor — also received Ian Osborne's assessment that "Peter is best for currency; maybe Reid Hoffman." Peter being Peter Thiel. Currency being cryptocurrency.
The same apartment building that housed Ehud Barak in unit 11J — with dedicated cleaning staff, alarm codes, Apple TV installation — also hosted a Heidi Klum "Hookers and Pimps" party email that named Ghislaine Maxwell's father.
Political. Financial. Intelligence. Academic. Media. Tech. Every network converged at one node. That's a switchboard.
CARVER
CARVER is a military targeting methodology developed by the U.S. Special Forces during Vietnam. It scores targets across six dimensions:
- **C**riticality — how important is the target to the network?
- **A**ccessibility — how exposed is the target to investigation?
- **R**ecuperability — how quickly can the network recover if this target is compromised?
- **V**ulnerability — how susceptible is the target to exposure?
- **E**ffect — what's the downstream impact of compromising this target?
- **R**ecognizability — how identifiable is this target to the public?
Each dimension scored 1-5. Maximum score: 30.
We scored 31 names. Not randomly. Each name was selected based on network position — how many connections to Epstein's inner circle, which networks they bridged, whether they appeared in the black book, flight logs, or victim testimony.
Some names had hundreds of documents in our index. Steve Bannon had 45. Larry Summers had 55. We scored them and moved on.
Seven names had zero.
The Zeros
Zero documents is supposed to mean "not there." That's how most people read an empty search result. You search for a name, you get nothing, you move on.
But zero isn't nothing when the network says otherwise.
Robert Maxwell — Ghislaine's father, confirmed Mossad agent, the man whose funeral was attended by six serving heads of intelligence — had zero documents in 136,000 pages released by the DOJ. The father of Epstein's primary co-conspirator. Zero pages.
We scored him 21/30.
Nicole Junkermann — whose investment fund received NHS Digital health data access after a UK government appointment, whose name appears in the Panama Papers, who was photographed with Epstein as early as 2002 — zero documents.
We scored her 19/30.
Sergey Brin — Google co-founder, 13 connections to Epstein's inner circle, photographed at the same events — zero documents.
We scored him 20/30.
Seven names. Zero documents. All scored on network position alone. The logic: if someone has 13 connections to the inner circle but zero documents in 136,000 pages, the documents aren't missing. They're somewhere you haven't looked yet.
That was the cipher. We published it. We waited.
Congress
On February 24, 2026, we indexed the House Oversight Committee's Epstein investigation files.
2,897 documents. Fundamentally different from the DOJ release. The DOJ gave us the investigation — what prosecutors chose to include in case files. Congress gave us the man's inbox. Emails sent and received. Personal correspondence. Scheduling. Financial records. Photographs.
Custodian: Jeffrey Epstein.
We parsed the Concordance load file — a metadata format used by litigation support companies, delimited by thorn characters and DC4 control codes, the kind of format that exists specifically because the legal industry doesn't want regular people to read it. We matched 2,897 metadata records to 2,895 extracted text files. We pushed them into the index.
Then we re-ran the cipher.
7 for 7
Robert Maxwell: 0 → 1,000+.
Nicole Junkermann: 0 → 677.
Jean Pigozzi: 0 → 496.
Peggy Siegal: 0 → 377.
Sergey Brin: 0 → 116.
Doug Band: 0 → 110.
Huma Abedin: 0 → 11.
Every zero-doc prediction hit. Every single one.
The probability of randomly guessing seven binary outcomes correctly is 1 in 128. But these weren't binary guesses — we scored them on a 30-point scale. We ranked them by predicted importance. And the ranking held: Maxwell (scored highest among zeros at 21/30) had the most documents. Abedin (scored 17/30) had the fewest. The ordinal predictions tracked.
The Spearman rank correlation between CARVER scores and House Oversight document counts was 0.319 — moderate, positive, real. Not r=0.95, which would suggest overfitting. Not r=0.01, which would suggest noise. The sweet spot where a model captures genuine signal but admits it doesn't explain everything.
The cipher worked.
What Congress Proved
The House Oversight data didn't just fill the zeros. It exploded the low-doc names.
John Brockman — the Edge Foundation literary agent who ran the intellectual salon circuit — went from 9 documents to over 1,000. A 111x increase. He'd invited Epstein to the Edge Annual Question in 2012, embedding a convicted sex offender into the world's most exclusive gathering of scientists and technologists for seven years.
Jeff Bezos went from 10 to over 1,000. A 100x increase.
Steve Bannon went from 45 to over 1,000. And the House Oversight data surfaced something the DOJ release never showed: Epstein self-emailed a document titled "list for bannon steve" on June 30, 2019 — two weeks before his arrest. The list contained names: Mandelson, Barnaby, Summers Watson, Axel Bottstein, Karim Terje. Epstein was curating intelligence target lists for Trump's former chief strategist.
Ehud Barak — Israel's former Prime Minister, who we'd already documented as having a dedicated apartment in Epstein's building — emailed Epstein about the Iran nuclear deal on April 23, 2015, mentioning Bill Clinton in the body. This was during the final months of the Obama administration's JCPOA negotiations. A former head of state was using a convicted sex offender as a diplomatic back-channel during the most sensitive nuclear negotiation of the decade.
Boris Nikolic — Bill Gates' science advisor, named as a backup executor in Epstein's will — emailed Epstein: "mission accomplished for some." January 13, 2014. The Gates Foundation's representative was reporting mission status to Jeffrey Epstein.
Michael Wolff — Fire and Fury author — was in direct email correspondence with Epstein through January 2019. Six months before the arrest. Epstein sent Wolff a Scottish Sun link about "teen sex Prince Andrew." The journalist and the sex offender were trading kompromat about a royal.
Before the House Oversight data: 6,536 total CARVER name hits across 31 targets.
After: 21,755.
+15,219 new name hits from 2,897 documents. That's 5.3 hits per document — the Congressional data is disproportionately rich because it's personal correspondence, not redacted case files.
31 out of 31 targets appeared. Every single one.
Why It Works
The cipher works for the same reason signals intelligence has worked since Bletchley Park: you don't need to read the message to know it matters. You need to see who sends it and who receives it.
Network topology is evidence. If a node has 13 edges to the inner ring but zero documents in the disclosure, there are only two explanations: either those 13 relationships were maintained entirely verbally across years and continents (no), or the documents exist and weren't included in the release.
The DOJ chose what to release. Their choices formed a pattern. The pattern had holes. The holes had shapes. And the shapes matched the names we scored highest.
That's the cipher. It doesn't predict guilt. It predicts where to find evidence. It maps the topology of suppression — what the government chose to hide, revealed by the shape of what they chose to show.
And it was 100% accurate.
The Part That Matters
Today — February 26, 2026 — Prince Andrew has been arrested in the UK. Peter Mandelson has been arrested in the UK. Thorbjorn Jagland, former Prime Minister of Norway, has been charged with aggravated corruption and is now hospitalized. Bill Gates confessed at a Foundation town hall. Kathryn Ruemmler resigned from Goldman Sachs. Tom Pritzker stepped down from Hyatt Hotels.
Six people. Twenty-six days. The cipher predicted where the evidence was hiding. Congress confirmed it. The searchable index delivered the consequences.
Marina Hyde wrote in The Guardian: "So Epstein buddies Andrew and Mandelson have been arrested in the UK. And in the US? Zero, zip, nada."
She's right. The UK has made two arrests. France charged their nationals. Norway charged a former prime minister. Ten countries have launched criminal investigations from these files. The United States — which produced the files, which conducted the investigation, which held the grand jury, which signed the plea deals, which ran the prison where Epstein died — has arrested no one.
The Tradecraft Beneath
The cipher predicts where evidence is hidden. But the evidence itself reveals how the network operated — and the tradecraft is intelligence-grade.
**Five email identities** — jeevacation, jeeitunes, littlestjeff, jeffreyepsteinorg, jeeproject — each compartmented for a different function. The FBI's own keyword list (EFTA00027057) documents all of them.
**Draft folder dead drops** — 3,475 documents reference the technique. No email is sent. One person writes in Drafts, another reads and deletes. No transmission metadata. Same technique used by CIA officer Jeffrey Sterling, the Anna Chapman Russian spy ring, and General Petraeus. Who taught a college dropout counterintelligence tradecraft?
**Coded vocabulary** — "massage therapist" appears in 48,115 documents. "Students" for victims in transit. "Assets" in logistics memos. A full vocabulary of concealment operating in plain sight for two decades.
**BlackBerry standardization** — 14,677 documents. BBM encryption, PIN-to-PIN messaging that bypassed carrier logs. Even Lord Mandelson was on BlackBerry when emailing Epstein.
**30 DNA kits to Dubai** — Epstein ordered 30 23andMe kits, had them delivered to his NYC house, and coordinated shipping to JFK for a flight to Dubai (EFTA02216142). Staff emailed 23andMe customer service asking if kits could be used internationally and returned for processing. A kit was dispatched by name to "Aziza." The 23andMe founder, Anne Wojcicki, attended Epstein's dinners alongside Yuri Milner — whose money Epstein described as "russian gangster money" (EFTA01882985).
**Maxwell's parallel network** — The FBI keyword list includes *gmax*, *pipex*, *aace*, *abx17*. Ghislaine Maxwell ran her own communication infrastructure. Her Terramar Project maintained separate UBS accounts at 116 East 65th Street, blocks from Epstein's townhouse.
This wasn't a rich man's indiscretion. This was an intelligence operation with compartmented communications, coded language, multi-jurisdictional safe houses, shell company layering, surveillance infrastructure, and DNA collection programs. The question isn't what Epstein did. It's who built the architecture for him.
The Confirmed Kills
We track outcomes at [epstein.dugganusa.com/medusa-statues.html](https://epstein.dugganusa.com/medusa-statues.html). Six stone busts. Each scored by CARVER before the outcome. Each confirmed by real-world consequence:
1. **Bill Gates** — CARVER 27/30. 6,134 documents. Confessed Feb 25, same day we published the Mila Antonova deep dive.
2. **Prince Andrew** — CARVER 27/30. 14,858 documents. Arrested Feb 19. First British royal arrested in 379 years.
3. **Peter Mandelson** — CARVER 25/30. 515 documents. Arrested Feb 23. Misconduct in public office.
4. **Thorbjorn Jagland** — CARVER 24/30. 407 documents. Charged Feb 12. Hospitalized Feb 25.
5. **Kathryn Ruemmler** — CARVER 22/30. 439 documents. Resigned Goldman Sachs Feb 12.
6. **Tom Pritzker** — CARVER 22/30. 5,637 documents. Resigned Hyatt Hotels.
Plus Larry Summers (resigned Harvard + OpenAI), Brad Karp (resigned Paul Weiss), Mona Juul (resigned Norwegian ambassadorship), and more still falling.
The Corpus
364,000+ documents. 31 CARVER targets. 7 zero-doc predictions validated. 6 confirmed kills. All government-sourced. All legal. All searchable at [epstein.dugganusa.com](https://epstein.dugganusa.com).
The CARVER matrix is published. The methodology is open. The cipher is falsifiable — take the scoring framework, apply it to the next dataset release, and see if the predictions hold. We believe they will. We're 95% confident. The 5% we're missing is the part that matters most.
One company. Two people. $500 a month. And the government's own evidence, made searchable, continues to indict the government.
*Sources: House Oversight Committee Epstein Investigation files (3 releases), DOJ EFTA Datasets 1-12, ICIJ Offshore Entities Database, Published CARVER Target Analysis Matrix*
*All data sourced from DOJ public releases. No leaks, no hacks, no felonies. Government-released documents, made searchable.*
*DugganUSA LLC — protect. publish. amplify.*
*Her name was Renee Nicole Good.*
*His name was Alex Jeffery Pretti.*
Comments