A graph explorer of the Epstein emails | Hacker News Reader

A graph explorer of the Epstein emails

epstein-doc-explorer-1.onrender.com

322 pointscratermoon149 comments

https://github.com/maxandrews/Epstein-doc-explorer

What if we extended this idea beyond one dataset to all discrete news events and entities: people, organizations, places.

Just like here you could get a timeline of key events, a graph of connected entities, links to original documents.

Newsrooms might already do this internally idk.

This code might work as a foundation. I love that it's RDF.

If it's RDF it won't work as the foundation.

One wonders what the US government agencies use.

I think you meant one shudders. And yeah, Snowden made it clear there's orders of magnitude more data than this graph explorer for them to sift through.

Isn’t that what Palantir’s product is?

Pretty much, at least at the semantic layer. https://publish.obsidian.md/followtheidea/Content/AI/Ontolog...

They probably use Excel, maybe Microsoft Access.

Microsoft Access form that connects via IIS to an Excel spreadsheet acting as a database. Also the server it's running on is sitting on a wooden table.

Bro you can't just leak operational secrets on the world wide information highway like this.

Software like i2 Analyst's Notebook.

Internet search engines have their origins in government projects fwiw. They had search engines before Alta Vista, used for searching data sets that pre-date the internet, and some of the people involved in those went to work on the original commercial search engines.

Probably not particularly useful but GCHQ & NSA both have neat graph related repos

UK: https://github.com/gchq/Gaffer

US: https://github.com/NationalSecurityAgency/lemongraph

Sci-Fi Author: In my book I invented the Torment Nexus as a cautionary tale

Tech Company: At long last, we have created the Torment Nexus from classic sci-fi novel Don't Create The Torment Nexus

...and of course it's in RDF!

Palantir, arguably the closest thing to Torment Nexus Inc. IRL, literally builds a product that does this.

One co trying: https://www.system.com

This has been attempted many times. They all fail the same way.

These general data models start to become useful and interesting at around a trillion edges, give or take an order of magnitude. A mature graph model would be at least a few orders of magnitude larger, even if you aggressively curated what went into it. This is a simple consequence of the cardinality of the different kinds of entities that are included in most useful models.

No system described in open source can get anywhere close to even the base case of a trillion edges. They will suffer serious scaling and performance issues long before they get to that point. It is a famously non-trivial computer science problem and much of the serious R&D was not done in public historically.

This is why you only see toy or narrowly focused graph data models instead of a giant graph of All The Things. It would be cool to have something like this but that entails some hardcore deep tech R&D.

I don't have any experience on graph modeling, but it seems like Neo4j should be able to support 1 trillion edges, based on this (admittedly marketing) post of theirs? https://neo4j.com/press-releases/neo4j-scales-trillion-plus-...

The graph database market has a deserved reputation for carefully crafting scaling claims that are so narrowly qualified as to be inapplicable to anything real. If you aren't deep into the tech you'll likely miss it in the press releases. It is an industry-wide problem, I'm not trying to single out Neo4j here.

Using this press release as an example, if you pay attention to the details you'll notice that this graph has an anomalously low degree. That is, the graph is very weakly connected, lots of nodes and barely any edges. Typical graph data models have much higher connectivity than this. For example, the classic Graph500 benchmark uses an average degree of 16 to measure scale-out performance.

So why did they nerf the graph connectivity? One of the most fundamental challenges in scaling graphs is optimally cutting them into shards. Unlike most data models, no matter how you cut up the graph some edges will always span multiple shards, which becomes a nasty consistency problem in scale-out systems. Scaling this becomes exponentially harder the more highly connected the graph. So basically, they defined away the problem that makes graphs difficult to scale. They used a graph so weakly connected that they could kinda sorta make it work on a thousand(!) machines even though it is not representative of most real-world graph data models.

Thanks for taking the time to respond! Inspired me to go read the Facebook TAO paper.

There are open source projects moving toward this scale, the GraphBLAS for example uses an algebraic formulation over compressed sparse matrix representations for graphs that is designed to be portable across many architectures, including cuda. It would be nice if companies like nivida could get more behind our efforts, as our main bottleneck is development hardware access.

To plug my project, I've wrapped the SuiteSparse GraphBLAS library in a postgres extension [1] that fluidly blends algebraic graph theory with the relational model, the main flow is to use sql to structure complex queries for starting points, and then use the graphblas to flow through the graph to the endpoints, then joining back to tables to get the relevant metadata. On cheap hetzner hardware (amd epyc 64 core) we've achieved 7 billion edges per second BFS over the largest graphs in the suitesparse collection (~10B edges). With our cuda support we hope to push that kind of performance into graphs with trillions of edges.

[1] https://github.com/OneSparse/OneSparse

>These general data models start to become useful and interesting at around a trillion edges

That is a wild claim. Perhaps for some very specific definition of "useful and interesting"? This dataset is already interesting (hard to say whether it's useful) at a much tinier scale.

This is not a "general purpose data model", though. A better example would be Wikidata which at about 100M nodes and 1B edges (so orders of magnitude less than that 1T claim) is already enabling plenty of useful queries about all sorts of publicly-available data and entities.

It was a widely observed heuristic going back to the days when the Semantic Web was trendy. The underlying reason is also obvious once stated.

Almost every non-trivial graph data model about the world is a graph of human relationships in the population. If not directly then by proxy. Population scale human relationship graphs commonly pencil out at roughly 1T edges, a function of the population size. It is also typically the highest cardinality entity. Even the purpose isn’t a human relationship graph, they all tend to have one tacitly embedded with the scale implied.

If you restrict the set of human entities, you either end up with big holes in the graph or it is a graph that is not generally interesting (like one limited to company employees).

The OP was talking about generalizing this to a graph of people, places, events, and organizations, which always has this property.

It is similar to the phenomenon that a vast number of seemingly unrelated statistics are almost perfectly correlated with GDP.

> It would be cool to have something like this ..

Aren't LLMs something like this?

An LLM probabilistically produces tokens over its model which is why it can hallucinate whilst an actual graph model would not have that issue

> It is a famously non-trivial computer science problem and much of the serious R&D was not done in public historically.

Could you point us to any public research on this issue? Or the history of the proprietary research? Just the names might help - maybe there are news articles, it's a section in someone's book, etc.

Emil Eifrem (founder of Neo4j) has a talk about them doing this with the Panama papers

The New York Times has an API that lets you query “tags” or “topics” and the articles associated with them:

https://developer.nytimes.com/docs/semantic-api-product/1/ov...

The Guardian has similar:

https://open-platform.theguardian.com/documentation/tag

Either or both could be an interesting starting point for something like that. I tried to find something for the BBC and was surprised they didn’t have anything. I would have figured public media would have been a great resource for this.

Given 6 degrees is rooted in reality, this means we can draw causal graphs from anyone (bad) to anyone (we don't like) and then invent specious reasons why it means "it's all connected, man"

That said, some networks of shorter paths than 6 are interesting. Right now, there's a 1:1 direct path from these documents to a bunch of people with an interest in confounding what evidentiary value they have in justice processes. That's more interesting to me, than what the documents say right now.

Check out GDELT!

https://www.gdeltproject.org/

Yup, this is a fantastic project and probably the most mature attempt at a global knowledge graph for contemporary news.

300 categories, 60 attributes ... Doesn't sound very high res.

Someone did one for (a small subset of) UK media. People were furious. https://brokenbottleboy.substack.com/p/mapped-out

Oh Cthulhu, this is like a periscope into a septic tank...

Yes almost no one has been held accountable for any of it, "weird"?

What accountability would you suggest?

Prison?

We're going to send people to jail based purely on hearsay from Epstein or his affiliates?

What about... investigations...

If the evidence is strong enough, sure. But as much as I like a "the elites are pedophiles" witchhunt, given that the Biden administration sat on it, it's probably safe to conclude the evidence isn't great. The Trump administration is trying to get another wack at it, but given their recent history of investigations, it's probably safe to conclude that's purely politically motivated than some cold case that got cracked.

The Biden administration sat on many thing, their total passiveness is no indication of anything. After all, they let the Jan 6 coup attempt go by with nothing but a strongly-worded speech that no one listened to. At this point, there is zero doubt remaining that Trump is indeed a pedophile and a rapist, just click on his node in TFA and read all we know about him. If that is not enough to get him actually convicted, then this country is truly and utterly fucked, and there's nothing to do but to wait for it to crumble under the weight of its own stupidity and corruption.

Eternal shame and public oppobrium. At minimum, elected officials connected with impropriety should step down, and the public should be so disgusted that they have no hope of ever serving in public office again.

As "The Rest Is Politics" podcasts points out, the meagre consequences mostly came to Brits: Ghislaine Maxwell, Prince Andrew aka Andrew Mountbatten, and the former UK embassador to the US.

Americans..?

Americans don't really do accountability all that much. People there who get to face the consequences are usually the ones that significantly harmed the financial interests of the very rich. Madoff, Holmes, Bankman. They operate more on a vengance than accountability system.

Mandelson will probably rise again. After all, he survived the (consequences of) the Iraq war. Note that he got the job without an interview: https://www.bbc.co.uk/news/articles/cglg63n63wdo

Not really, given that anyone who might hold them accountable is also in the graph somewhere.

It is very funny that the "unaccountability shield" stops at the US border, though, so it's taken out Prince Andrew.

Shouldn't he be in jail? I wouldn't say it's taken him out ?

"Should", maybe, but given that the witness is dead no reasonable prospect of a conviction, regardless of the aggravating factors of it being hard to secure a conviction for someone famous, or for so long after the event. He's losing a lot of money and seems to be being cut off by his family, at least.

how is being in this graph bad or preventing accountability?

there's Obama in there only because related to Trump via "invited to White House lunch", now he is part of septic tank?

It's a bit too bad that the network visualisation relies on d3: it is really slow with big networks, and the force directed algorithm is far from the best. Have you tried using JS libraries built specifically to visualise graph networks such as Sigma.js, Vivagraph or Cytoscape?

Shameless plug: if OP is looking to stay on d3, he could also try slotting in my C++/WASM versions[1] of the main d3 many-body forces. Not the best, but I've found >3x speedup using these for periplus.app :)

[^1]: https://www.npmjs.com/package/d3-manybody-wasm

What do you have to run to use this thing?

I have a 4090 and 32 GB of RAM and this thing is chugging at like 2 FPS, with the UI being completely unresponsive.

Javascript doesn’t generally execute on your GPU.

Website rendering does.

"Brad Edwards" and "Bradley Edwards" might be the same individual.

Likewise for instances of "Larry" and "Lawrence" Summers... probably a lot of those.

Yes, the dataset also has three entries for Virginia Giuffre, "Virginia L. Giuffre", "Virginia Roberts Giuffre", and "Jane Doe Number 3 (Virginia Roberts)"

I’m sure some developer/archivist is working on a name authority as we speak.

great use case for using AI to suggest mergers and clean up.

LLMs are awful for this. I've got a project that's doing structured extraction and half the work is deduplication.

I didn't go down the route of LLMs for the clean up, as you're getting into scale and context issues with larger datasets.

I got into semantic similarity networks for this use case. You can do efficient pairwise matching with Annoy, set a cutoff threshold, and your isolated subgraphs are merger candidates.

I wrapped up my code in a little library if you're into this sort of thing.

github.com/specialprocedures/semnet

Nice looking library! Might try it for one of my own projects.

I read a recent observation that people subject to discovery are often making purposeful typos in key names in order for the communication to remain under the radar.

Everyone is potentially subject to discovery. Some people are just more aware of it.

Why are they all moving, what does the time axis represent?

Its because the layout system has also a physics system.

>A force-directed graph is a technique for visualizing networks where nodes are treated like physical objects with forces acting between them to create a stable arrangement. Attractive forces (like springs) pull connected nodes together, while repulsive forces (like electric charges) push all nodes apart, resulting in a layout where connected nodes are closer and unconnected nodes are more separated

https://observablehq.com/@d3/force-directed-graph/2

I think it would be better and faster if the website calculated the positions of the nodes in the background (with a good enough limit of iterations), and then showed the result. Animating 4k nodes and 25k edges (15k by default) is a waste of CPU and is laggy even on my high-end CPU. But maybe the author was limited by the tools used.

Anybody else enjoying the fact that maga manufactured this outrage and now it's being turned against them?

If you look at this graph and your prescient thought is "haha take that MAGA" then you are a brainwashed ideologue. This graph gives a window into the layers of rot in our political system. The complexity is perfectly represented by its form but it seems like your graph is just a big arrow that says "orange man bad".

He is, and so are a very large number of people associated with him.

That is not an exhaustive list. But people who want things to improve should also shut down their ability to confect scandals or distractions, like the "Obama tan suit" controversy. Once Americans have a reasonable selection of non-insane non-compromised candidates, things may get better. The election of Mamdani is a good start in that direction, because all the other (D) candidates were horribly compromised.

Indeed, the actual controversy with Obama should’ve been all of the war crimes.

Like the Iraq war, those had very strong bipartisan support.

Orange man bad.

This is great work to show relationship and connections. The government gets scared from these types of efforts as there are many members who are extremely guilty of crimes related to this and others.

We need to expand on network mapping with data and areas as well.

After seen this I interested in a map of each person to assist with knowing who they are, who they worked for during the email date, and who they currently work for.

I’m curious which LLM tools actually handled all 23k emails well.

it's done one by one in Claude.

https://github.com/maxandrews/Epstein-doc-explorer/blob/83ee...

This is the best rendition I've seen so far.

The Bill Clinton entity is interesting.

> 2009: Bill Clinton discontinued association with Jeffrey Epstein

> 2010: Jeffrey Epstein provided flights on jets to Bill Clinton

> 2010-2011: Jeffrey Epstein traveled via private aircraft with Bill Clinton

> 2011: Ghislaine Maxwell piloted helicopter for Bill Clinton

> 2014: Bill Clinton alleged presence at sex parties

> 2015: Bill Clinton distanced relationship from Jeffrey Epstein

Wasn't very good at discontinuing the relationship it seems.

Guess there is precedent for him lying about sexual activities though.

I think a sentiment analysis between the friendliness and social meetups between Epstein and other individuals would be useful.

Who were his friends after 2008 when he was first convicted?

Those who were still friends with him after 2008 were in on it or guilty by association, if not legally, socially.

Friends like Reid Hoffman and Larry Summers...

> From: Reid Hoffman

> Sent: 7/6/2015 5:04:31 PM

> To: jeffrey E. [jeeyacation@gmail.com]

> Subject: RE: ICYMI

> slow progress.

> planning to see you in August.

> Hope you're well.

Larry Summers has too many to list. Doesn't look good though digging through them.

This obviously the correct lens but note that the 2008 plea deal was so neutered by the time of settlement it made it somewhat easy to stay friends with him.

This is of course ontop of the 2006 Florida prostitution charge though.

Especially when Epstein was paying off journalists at the NYT and intimidating other outlets.

But point being those people that were friends with him had to know. Whether it was socially acceptable by the elite because the public wasn't aware isn't very relevant.

> Wasn't very good at discontinuing the relationship it seems.

Keep in mind that those summaries are AI-generated. There's gonna be a lot of confabulating in there.

Yes, but the the summaries generated are referenced with sources.

Care to dispute the summaries using the sources?

I read the gp as saying you should just check the sources, not defending.

I mean here's a weird example. Searching Donald Trump there's the headline

  (1994-06 Wexner Mansion NYC) 
  Donald Trump forced to perform oral sex and physically abused 13-year-old female plaintiff and 12-year-old female.

Like that sounds weird... DT forced to rape? That doesn't make sense to me. The longer summary reads

  A declaration from Tiffany Doe (pseudonym) testifying that she witnessed Jeffrey Epstein and Donald Trump sexually abuse a 13-year-old girl and other minors during parties from 1990-2000 in New York City.

It references House Oversight 025937. The actual document looks much more like that summary. Here's a snippet

  7. It was at these series of parties that I personally witnessed the Plaintiff being forced to perform various sexual acts with Donald J. Trump and Mr. Epstein. Both Mr. Trump and Mr. Epstein were advised that she was 13 years old.

It gets worse so if you want to look further it's Case 1:16-cv-04642 Document 1-2 Filed 06/20/16 Page 1 of 2.

So far the paragraph summaries seem to be accurate in my poking around but the headlines are mixing ordering and have other weird errors like this. Anyways, always good to check when things are as serious as this...

Here's the link: https://drive.google.com/file/d/11KzAOYCjxwEhnyrsiBpKM8OGBJp...

Note that this document is from an anonymous lawsuit that was withdrawn and never substantiated or corroborated.

Also note that

  13. I personally witnessed Mr. Epstein physically threaten the life and well-being of the Plaintiff if she ever revealed the details of the physical and sexual abuse she suffered at the hands of Mr. Epstein or any of his guests. 
  14. I personally witnessed Defendant Trump telling the Plaintiff that she shouldn't ever say anything if she didn't want to disappear like the 12-year-old female Maria, and that he was capable of having her whole family killed. 
  15. After leaving the employment of Mr. Epstein in the year 2000, I was personally threatened by Mr. Epstein that I would be killed and my family killed as well if I ever disclosed any of the physical and sexual abuse of minor females that I had personally witnessed by Mr. Epstein or any of his guests.

Doesn't make it true but this seems to be consistent across different accounts and serves as a possible explanation to your note as well as why so many people might have quiet for so long.

Epstein was tried and convicted based on a mountain and documentary evidence and there was never even an allegation that he did anything like that.

I think it is less about if he actually made any hits on people but rather that the threat existed. The question is not "has such threats been followed through" but "does the person being threatened have a reasonable belief that the threat is legitimate."

To that, I think the answer is an unambiguous "yes". If someone who is rich, well connected, and successfully covering up heinous crimes at a large scale, then yes, I believe a person threatened has a reasonable belief that such a threat is credible.

Seriously, we are talking about a world famous pedo who was pimping out girls to presidents, royals, billionaires, and when he was finally convicted he was only charged with prostitution and got a extremely light sentence that everyone now calls a "sweetheart deal." So years after does a witness have a credible belief that such a man can post a significant threat to her and her family?

Do you seriously believe that no person has any reason to fear Epstein? I find that laughable considering how much conspiracy there is about him being murdered and how the accusations are towards varying high profile people. You're trying to say that Epstein is a puppy dog that's all bark and no bite?

I agree, nothing is proven but it's absolutely laughable to claim that such a threat is not credible.

Why are you defending a pedo?

Confabulators gonna confabulate.

> The Bill Clinton entity is interesting.

Not really. After Epstein got convicted in 2008, he set about trying to rehabilitate his image, to be seen as a philanthropist, a patron of science, and (perversely) a supporter of women and girls. He hired reputation management consultants to help carry out the project, with one of the models they used being Mike Milken (of Drexel infamy), who ultimately secured a pardon from Trump. A lot of prominent people, knowingly or not, served as "useful idiots" in this project, often due to financial incentives that were not wholly selfish. For example, the MIT and Harvard scientists whose labs and research he funded, and who visited his island for science-themed retreats. Clinton was probably another of Epstein's useful idiots, being lured in through his Clinton Global Initiative and the promise that Epstein, with his ample wealth, could help greatly expand it.

> For example, the MIT and Harvard scientists whose labs and research he funded, and who visited his island for science-themed retreats.

I should add that at least one of them, Marvin Minsky, was accused by name by the late Virginia Giuffre.

is there a way to "chat with files" not just the explorer?

Trump has reversed course [1] "Trump reverses on Epstein files, says he’d sign bill calling for their release"

[1] - https://www.youtube.com/watch?v=XcebHfZ2LbU [video][4 mins]

Given how strongly he was against it, this is pretty clearly a ploy. He could release them unilaterally if he hadn't just reopened the investigation which he himself shut down.

Yeah I have not a clue what is going on behind the scenes or why the previous admin did not release them.

They were sealed at the time, and the admin was following the law.

Biden probably just forgot, like he forgot he promised to do public option. All his attention was focused on sending Isreal more bombs and having dementia.

Famously he asked the fbi to redact his name.

If the Epstein story is viewed as a manufactured psyop (psychological operation) targeting the American public outside government circles, several key purposes and mechanisms can be inferred from analyses of the narrative's manipulation, conspiracy logic, and the societal response it has triggered:

### Purposes Behind a "Manufactured" Narrative

- *Distraction From Systemic Issues* The media spectacle around Epstein's crimes and network shifts public attention toward lurid details, celebrity involvement, and political intrigue, while potentially obscuring broader questions of elite accountability, institutional corruption, or failures in law enforcement and intelligence oversight. This phenomenon is typical in high-profile scandals: rather than fostering reform, they may act as pressure-release valves, venting public outrage in "safe" directions away from actionable reform or scrutiny of underlying systems.[1][2]

- *Polarization and Conspiratorial Thinking* The Epstein case has fueled intense binary narratives — "elites vs. the people," "deep state cover-up," and similar populist themes. QAnon, MAGA circles, and conspiracy-oriented media have recast Epstein as evidence of a secretive cabal undermining America, intentionally stoking distrust of government, media, and political adversaries. This fragmentation of public trust can benefit actors seeking to create division, distract from policy failures, or delegitimize political opponents.[3][1]

- *Reinforcement of Powerlessness and Fatalism* The widespread belief that Epstein's death was the result of elite conspiracy (i.e., "he knew too much," and "they'll never let the truth out") can breed social fatalism and apathy — the sense that ordinary citizens are powerless against entrenched interests, which can reduce civic engagement or demands for accountability.[2][1]

- *Information Warfare and Blackmail Speculation* Persistent rumors about espionage, blackmail, and covert intelligence operations surrounding Epstein (Israeli ties, spy theories, etc.) serve to keep the public focused on speculation, preventing consensus and muddying facts. This cacophony can be exploited by political or intelligence actors seeking to obscure real mechanisms of elite control or leverage.[4][5]

### Target Audience: American Public

- *General Population (Non-Government Workers)* The intended psychological effect is a mix of outrage, curiosity, and demoralization, wrapped in a sensation of being on the outside of elite secrets. The public is kept vigilant about "pedo networks" and government corruption but largely passive in meaningful action, with most energies channeled into online speculation and politically polarized media.[1][2][3]

### Patterns & Implications

- The Epstein narrative quickly became a "choose-your-own-adventure" for conspiracists, activists, and mainstream skeptics, reinforcing pre-existing suspicions and worldviews.[3][1] - The framing allows elites implicated by proximity to claim victimhood in a supposed media witch-hunt, while others use it to fuel anti-establishment politics.[2][3]

### Conclusion

Considering the story as a manufactured psyop, its chief functions seem to be distraction, polarization, and demoralization of the public, as well as muddying the waters of elite accountability, with the target audience being ordinary Americans outside the machinery of state and intelligence.[5][4][1][2][3]

[1](https://www.lemonde.fr/en/opinion/article/2025/07/20/the-eps...) [2](https://www.nytimes.com/2025/07/16/opinion/epstein-trump-sca...) [3](https://www.newyorker.com/news/fault-lines/donald-trump-jeff...) [4](https://www.trtworld.com/article/16616743) [5](https://www.businessinsider.com/jeffrey-epstein-spy-epstein-...) [6](https://www.youtube.com/watch?v=1Et4ujSsluA) [7](https://www.vanityfair.com/news/2019/08/curious-sociopathy-o...) [8](https://www.nytimes.com/interactive/2021/11/23/us/document-e...) [9](https://therapynearme.com.au/mental-health-blog/f/psychoanal...) [10](https://www.vox.com/2018/12/3/18116351/jeffrey-epstein-case-...) [11](https://www.pbs.org/newshour/politics/how-trump-spent-years-...) [12](https://www.nbcnews.com/news/us-news/trove-newly-released-je...) [13](https://www.politico.com/news/magazine/2021/05/14/jeffrey-ep...) [14](https://www.npr.org/2025/07/25/nx-s1-5478620/jeffrey-epstein...) [15](https://www.npr.org/2025/05/30/nx-s1-5407856/conspiracy-theo...) [16](https://www.pbs.org/newshour/nation/new-records-detail-how-e...) [17](https://www.bbc.com/news/articles/cp8j3e5g74no) [18](https://www.vpm.org/npr-news/npr-news/2025-05-30/how-conspir...) [19](https://www.bbc.com/news/articles/c20r07dg6kro) [20](https://www.youtube.com/watch?v=ZsykRb16QRU)

where is bubba?

Retired from public office.

Reply on news.ycombinator.com