Recently, the Epstein case files have been intensively unsealed, and social media is full of people digging up the list.

People in the crypto circle are also asking: Are there people from our industry involved? I’m also quite curious, so I’m preparing to investigate.

But the problem is that the amount of published documents is too large, thousands of PDFs, manual searching is not realistic, and I don’t have much coding experience. I did a lot of online surfing unintentionally until I found a project on GitHub.

This project is called Epstein Doc Explorer. The author used AI to structure the Epstein legal documents released by the House Oversight Committee, extracting the relationships between people, events, times, and locations, and storing them all in a database.

It also has an online version of a visual webpage for you to search, but because there is too much data and my computer is relatively basic, it is very slow after entering, and I can’t get the information I want. What I want is batch query, to pull out all the crypto-related people and things at once.

Crawling method

I threw this github project to Claude and asked it how to query this database. After Claude read the project structure, it told me that the core data is in an SQLite database file called document_analysis.db, and there is a table called rdf_triples. Each record is a structure of “who – did what – to whom – when – where”.

However, the entire database has more than 260,000 such records, covering all the publicly available relationship networks in the Epstein files, and I only want crypto-related ones.

At this time, my idea was to filter out relevant records using keywords. But I stepped on two pits in actual operation.

The first pit: the file is “fake”. I downloaded the project compressed package from GitHub, and after decompressing it, I tried to open the database, but the computer couldn’t read it at all. After Claude helped me troubleshoot, I found that what I downloaded was just an index file, not a real database. I went back to the GitHub page and manually downloaded the complete file, a total of 266MB.

The second pit: the choice of query words. In the first version of the query, I used 25 keywords, including direct words such as Bitcoin, Crypto, Blockchain, and also added not-completely-related names and organizations such as Peter Thiel, Bill Gates, Goldman Sachs. The idea is to rather see it wrong than let it go.

As a result, 1628 records were retrieved, with a lot of noise. For example, 90% of the results from searching Goldman Sachs were economic forecast reports, and the results from searching Virgin Islands were local tourism data.

So I asked Claude to help me narrow down the search in three rounds:

First round: Broad keywords, 1628 entries, a lot of noise but established a panoramic view, locking in a group of people

Second round: After Claude read the results of the first round, he found that I didn’t search for some exclusive terms in the crypto field, such as Libra, Stablecoin, Digital Asset. After adding them, run it again to confirm that there are no omissions.

Third round: Search in reverse. Don’t search for names, only search for records containing Bitcoin, Crypto, Blockchain in “what did you do”

The third round is the most critical, directly filtering by behavior, and every record retrieved is strongly related. After cross-comparison of the three rounds, the weakly related ones were removed, and the following list was compiled.

1. Brock Pierce: Demonstrated Bitcoin to him at Epstein’s restaurant

Identity: Co-founder of Tether, Chairman of Bitcoin Foundation, Co-founder of Blockchain Capital

Contact with Epstein: Multiple face-to-face meetings at Epstein’s mansion on the Upper East Side of Manhattan.

Crypto topics involved: Bitcoin demonstration, blockchain discussion, cryptocurrency volatility.

Timeline: The earliest record is marked as 2008, and many records have no specific date but the location is the same. There are email records in March 2015 showing that someone arranged a follow-up meeting.

The reason why I took it out to write first is because Pierce’s record has the strongest sense of picture in the entire database.

The record shows that he gave a Bitcoin presentation in the restaurant of Epstein’s mansion in front of former US Treasury Secretary Larry Summers. After listening, Summers expressed concern about investment risks, but also “provided opportunities” to Pierce.

Epstein was not just listening in the side, he took the initiative to call Pierce to the front hall for a separate conversation, and then invited him back to the restaurant to continue chatting. The two also discussed the volatility of cryptocurrencies separately.

This is not an accidental dinner chat.

An email on March 31, 2015 shows that a person named Alex Yablon specifically wrote to Epstein asking if he could arrange for Pierce and Summers to discuss Bitcoin.

Someone is arranging the meeting, indicating that this type of meeting is deliberately organized.

Epstein’s Manhattan mansion played the role of an informal roadshow venue during that period. Pierce brought the project to the door, and the guest was the former Secretary of the Treasury.

For a crypto entrepreneur, this kind of resource channel hardly exists in normal commercial channels. Conversely, by arranging such meetings, Epstein put himself at the connection point between the crypto industry and policymakers.

Our database only processes the documents of the House Oversight Committee, and the Pierce clues that can be pieced together end here. But just as this article was being written, the US Department of Justice released another batch of more than 3 million pages of Epstein documents, and Decrypt made it based on these new documents

[深潮TechFlow]