Michael Wolff

Named on the documents as	”Michael Wolff” in the From: and To: lines of his own e-mail; “Michael” in the salutations of the assistants who booked him; “m” and “mw” in his own typed sign-offs; “Michael Wolff” in a dinner guest list and on Epstein day sheets123456
Titles the documents give him	None. The sheets read at the image for this page give him no title, employer or affiliation. What they show him doing is writing to Jeffrey Epstein about the press, and being booked into Epstein’s diary16
What this page is about	What the release records him writing to Epstein and to Epstein’s office between 2011 and 2019: a proposal to rebuild Epstein’s public identity, advice on how to answer reporters, offers to make introductions, and comment on drafts. It also sets out, at each point, what those documents do not establish
The three documents the desk commissioned this page on	EFTA01791807 (3 January 2015, press response); EFTA01987731 (4–5 October 2011, a proposed reputation project); HOUSE_OVERSIGHT_032229 (16 December 2015, a proposed debate question). All three record advice. None records that the advice was taken, or that anything followed from it1789
The project, in the record’s own sequence	Proposed June 2011 · discussed October 2011 · costed and declined-to-cost November 2011 · an outline still being asked for in August 2013 · still outstanding in September 201310731112
Documents read at the image for this page	Seventy-nine sheets across fifty-six documents, rendered locally, at 400 dpi except where a footnote says otherwise; ten sheets read at 600 dpi to settle two single letters, a third letter the sheet does not settle, the placing of the soft breaks on three copies of one message, the order of a nesting, a redaction bar and four damage claims, and fourteen sheets read at 300 dpi where the only question was whether any message stands on them
How the clocks on these sheets work	The unlabelled Sent: field in this export is coordinated universal time. Seven sheets read at the image carry an Apple date-received integer beside that field and six of them are relied on here: on five of those six the integer and the field agree to the minute in UTC, on the sixth the integer falls twenty-nine seconds short of the minute the field prints, and the seventh is ninety seconds off and is not relied on. A separate check runs the other way: one sheet’s field matches a sibling copy’s explicit +0000 header13
Not established here	That any advice on these sheets was acted on; that the reputation project was ever carried out; that the three introductions the record does show being made led to anything further; that any scheduled meeting took place, except where a later message says so; the identity of any person a sheet bars; or any characterisation of his role beyond what the documents themselves say

Michael Wolff is the journalist and author whose correspondence with Jeffrey Epstein is one of the larger single-correspondent files in the Epstein Files Transparency Act release: the exact phrase "Michael Wolff" returns 5,620 pages in 5,388 documents, and the union with ten damaged spellings accepted after sampling returns 5,648 pages in 5,413 documents.¹⁴ Most of that volume is duplication, book scans and press text. What is left, and what this page is about, is a run of messages in his own name: a plan to rebuild Epstein’s public identity, put to him in 2011 and still unfinished in 2013; advice on how Epstein should answer reporters; offers to put him in touch with people; and running commentary on the media as Epstein’s case moved back into the news.

Every section below keeps two things apart: what he wrote, and what happened. The distinction is not a formality here. The documents in this release are overwhelmingly messages in which somebody proposes something; the release very rarely records the next step. A message recommending a course of action is evidence that the recommendation was made. It is not evidence that it was followed, and this page does not treat it as such. Where the record does carry a later document bearing on whether something was done (and on the reputation project it carries five, over twenty-seven months), that document is set out too, in whichever direction it runs.

Nothing on this page is a finding of wrongdoing. He is a working journalist, and advising a subject on how to handle press coverage is not in itself an improper act; it is what the documents show, set out in the order they were written.

2011: press advice

The earliest sheet opened for this page is from July 2011, and it is already advice about a reporter.

On 20 July 2011 Epstein wrote: “are you around this weekend,, tina has asked Wayne Barrett to do an investgative piece. ? how do i appease her. this is crazy”. At a Sent: field of 5:48 PM on the same day Wolff answered:

Good news is that nobody takes Wayne Barrett seriously. I would certainly suggest not cooperating with him in any way. Worst that will happen is that he rehashes old stuff—exactly why we have to start building a counter narrative. Then, even if only begrudgingly, any story would have to include what you are doing now. Let me think about Tina, she seems to have it in for you and there are lots of way to neutralize her. I’m leaving for London on Friday night. Can you talk tomorrow?

The sheet gives the first name “tina” and no surname, and this page supplies none.¹⁵ The reporter it names is the one whose 2011 questions to the Governor of the U.S. Virgin Islands are set out on Paul Hoffman’s page; a separate August 2011 thread in which Epstein forwarded those questions to Ian Osborne and to Wolff under the line “suggestions?” is cited there.

2011 to 2013: the reputation project, and the outline

The second anchor the desk named is one sheet in the middle of a strand that runs for more than two years. Read on its own it is a proposal; read with the four other exchanges the release holds, it is a proposal that was still being chased twenty-three months later.

June 2011. At Mon, 13 Jun 2011 14:25:47 +0000, answering Epstein’s two-word message “wed thurs?”, Wolff wrote in full: “Ian has promised me he will have outline of a plan to you today. He’ll be in NYC tomorrow morning and we can meet together. Or I can meet on Wed morning—I’m leave for St. Petersberg on Wed at 6pm. m”¹⁰ Two days later Osborne wrote to Epstein, copying Wolff, under the subject “Meeting”: “Jeffrey, / Good to see you yesterday. We should meet again once you’ve had a chance to review my thoughts — copy attached.”¹⁶

October 2011. At 7:12 AM on 4 October 2011, in the quoted line the thread preserves, Epstein wrote: “I am aware that it might be time to start redoing the pages. are there concrete things that you suggest . I have been doing as little publicly as possible”. Wolff’s reply, which carries an explicit Date: Tue, 04 Oct 2011 12:57:18 +0000 header on the copy in the ninth data set, reads in full:

Should begin with strategy, goals, and story and then move into design and structure. It’s a good process that makes you think through both identity and message. Let’s discuss steps.

¹⁷ At 04:19:44 +0000 on 5 October 2011 Osborne answered both men: “Michael is right. Should Michael and I circulate an outline and then have a call later this week to discuss? / In the interests of time we should aim to do sooner rather than later. As part of redoing the pages, we also need to focus on Google.”⁷¹⁸ Seven hours later, at 11:22:52 +0000, Wolff added a third message: “Times story today (http://tinyurl.com/43enzct) on Amanda Knox is another good example of the importance and effectiveness of overall web strategy.”¹⁹

November 2011, when it came to money. On 7 November Epstein wrote “why dont you send ,me a financial proposal.” The answer, quoted on the sheet as sent at 11:19 AM, is the fullest statement in the record of what was being proposed:

The question is how much you want to do. In my estimation, the full plan, rolled out over 18-24 months, has a transformative effect and would allow you to be as public as you want to be, and, I believe, enable you to be who you want to be without any meaningful blowback. This allows us to not only take advantage of opportunities as the arise, but allows us to create opportunities through controlling your own story—putting you in a position to be as aggressive, creative, controversial, or authoritative with your public identity as you might want. You would be immunized from opprobrium and so would the people who associate themselves with you. Obviously, we could take a lesser, piecemeal course too. We could begin by just developing a new web presence—both a news site and a clean-up of the existing search results. A good part of this could be done without substantial time from you—at least in the initial backend phase. Then, too, if you’re still unsure about what you want to do, I have no problem continuing to chat with you about it on a casual basis. mw

Epstein came back: “understood, what do you envision as the time and money , in the various categories. or are you suggesting a full time job, or ghost writer, or…”. At Tue, 08 Nov 2011 10:37:17 +0000 Wolff answered: “fulltime job, perish the thought. I thought Ian had sent you an outline of various possibilities—since I am allergic to all money discussions. Let me chat with him today.”³

That sentence is a belief, not a fact, and the distinction is the whole of what the release settles here. Wolff wrote that he thought an outline had been sent. No outline is among the documents any search set out below returns, and twenty-one months later Epstein was still asking for one.

August 2013. On 19 August, under the subject “Some thoughts”, Wolff sent a lettered memo. It opens “An interesting predicament…” and runs:

A) The status-quo is to your liking;

B) By advising Gates, you run the risk of upending the status quo, causing both upset to yourself and to Gates. Any connection of you to Gates generates at minimum eyebrow raising coverage, aimed at Gates as much as you, or, possibly, a shit storm. The connection might not be made, but also might be made at any time; chances of the connection being made rise proportionally with your continuing involvement with Gates.

C) You can bet on the possibility of not being connected and hedge by putting crisis people in place and having them ready to respond if and when the connection is made—much goes awry in these situations in the gap between exposure and getting competent people to deal with the exposure.

D) You can out yourself. You can do it in a way that tries to preserve as much as possible of the status quo—for instance, Gates referencing you in some way and saying something positive about the importance of your contribution; or you giving an interview to a dull financial journalist, with positive comments from Gates, i.e. not hiding anything, but putting it on the record as undramatically as possible. But it is also likely that this will just provide the opening for the shit storm anyway. On the other hand, you can try to define a new normal by redefining yourself—the JE we don’t know—someone different enough from the past press that the story becomes that much more complicated and, hence, that much more difficult to slap a label on.

E) Unless you are luckier than you should plan on being, the status-quo is lost.

F) Is the Gates involvement worth that?

Then the offer:

Here is one unique approach which I think I can offer: I believe that I could write a redefined Jeffrey Epstein story—personal, fond, complex, funny, smart—that would deal with all the issues in a nuanced enough way that might make it anti-climactic for anybody else to try to deal with them. We’d seize the narrative. I’d be pleased to do this not just because I think it would help the situation, but also because I think it would be fun to tell your story. I’d suggest doing this sort of thing in New York Magazine. This approach involves more personal exposure but it seems safer to me than trying to low ball the story—again, that’s just hoping you don’t get noticed.

He added: “In addition, I would certainly put people in place now who are ready to deal with any sort of eruption of the story”, “Then there is long term strategy for after this is out in the open. That has to be thought through”, “Anyway, in a nutshell”, and “I leave for Berlin on Sunday and then London. I’m back on September 7.” At Mon, 19 Aug 2013 23:50:38 +0000 Epstein replied in one line: “when you are away , why don’t you think of an outline so we can discuss”.¹¹ Wolff assented. At Tue, 20 Aug 2013 00:51:07 +0000, one hour and twenty-nine seconds later measured between the two Date: headers, the whole of his reply is “my pleasure”.²⁰ The surname “Gates” is printed bare on the memo sheet eight times and is given no forename, and this page supplies no reading of it.

September 2013. On 24 September at a quoted 4:42 PM Epstein wrote: “you didn;t get a chance to share your outline ?” At Wed, 25 Sep 2013 10:44:28 +0000 Wolff answered, and the whole message is one line: “Finishing up something to today, but on the job. I thought you looked handsome in the Post.”¹²

What this strand establishes, and what it does not. A reputation project was proposed in June 2011, described in October, priced in principle in November, and returned to in August 2013 with a concrete offer of a magazine piece. Twenty-seven months after the first promise of an outline, the outline was still outstanding, and the last word on it in the record is “on the job”. No outline, no design, no site, no invoice and no contract reaches this page. The disconfirming document would be any of those five, and the search that would find one is set out below: "Michael Wolff" together with outline returns nineteen pages, and every one of them resolves to one of the five exchanges above: not one of them is an outline.¹⁴

A sequence is not a causal chain, and this page asserts none. The August 2013 memo proposes a piece in New York Magazine; a draft, a fact-checking list and a portrait shoot for that magazine are in the record in 2015. Those are the facts, in that order. Nothing on the sheets read here says the second followed from the first.

3 January 2015: “today’s papers”

The first anchor is a single Saturday’s exchange, and it survives across three production numbers whose page images between them carry every message in it.

It begins outside it. At 5:38 AM Osborne sent Epstein a message headed “today’s papers”: “Jeffrey, / Happy New Year. / =div>Rather unfortunately, your ongoing case has made the front page of eve=y British newspaper today (see attached). / I hope =hat all ok. / Best wishes, / Ian”.²¹ A second copy of the same forwarded message, produced under a different number, carries the soft breaks in different places and prints “Happy New Yea=.”, “has mad= the front page of every” and “Best wis=es,”.²² Epstein forwarded it to Wolff at 6:19 AM with one word: “whoops”.²¹

What followed. The nesting on these sheets is branched and is not a time order, so the messages are set out here in the order of the clocks the quoted headers give, each clock as its own sheet prints it:

• 9:55 AM, Epstein: “kathy ruemmler former white house =ounsel and good buddy , is suggesting she responds on my behalf to the pre=s inquires. ? thoughts?”¹
• 10:08 AM, Epstein: “if she is the spokesperson it will ratchet u= the attention thoughts/?”²³
• 10:10 AM, Wolff: “Probably. Issue is how much this=is spreading to US, or if it is still remains mostly British tabloid story=”²³
• 10:14 AM, Wolff: “Can you give me time middle of the month and=schedule Gates? Obviously, no time like now to deal with this. And this wi=l be a game-changing rebuttal.”²³ As in the 2013 memo, the sheet says only “Gates”, and this page supplies no reading. This one message survives under three production numbers, and each copy breaks its soft lines in a different place, which is why each damaged form below is pinned to the sheet that prints it: EFTA01791592 p.1 prints “and=schedule Gates” and “this wi=l be a game-changing rebuttal”;²³ EFTA01790536 p.1 prints “and schedule Gate=” and “a game-c=anging rebuttal”;²² and EFTA01791807 p.2 prints “and schedule Gat=s” and “a game-=hanging rebuttal”.²¹ All three were read at the image at 600 dpi. A string that verifies against some copy of a message is not a pin that is right, and an automated check of quotations cannot see the difference, because every one of these strings exists somewhere in the run.
• 11:03 AM, Wolff: “I think that is a reasonably=good approach. I might make the central point that court documents are a b=ck door way to make protected libelous allegations. Plaintiffs in pursuit =f settlements can wildly and improbably allege anything they want—the med=a then can repeat this without any obligation to verify. I would add this =s particular true of the British tabloid press, and in particular, the Dai=y Mail and its campaign against Prince Andrew. Then issue the emphatic den=al.”²³
• 11:38 AM, Wolff: “A statement from you or your reps does open in up in the U=S. Is the Times or WSJ calling you for comment?” The clock is the anchor document’s, which quotes the message as “On Sat, Jan 3, 2015 at 11:38 AM, Michael Wolff” and prints the wording differently again, as “A statement from you or=your reps does open in up in the U.S. Is the Times or WSJ calling you for =omment?”¹ The wording given first is as EFTA01791592 p.1 prints it, where the same text stands as the body of that sheet’s own 3:39 PM message and no clock of 11:38 appears at all.²³
• Epstein: “not yet. bu= they will,”¹

The order above is the order of the clocks, and it is not the order of the sheet. Read top to bottom, EFTA01791592 p. 1 carries the body of its own 3:39 PM message, then the 10:10 message with the 10:08 one quoted inside it, then the 11:03 message, then the 9:55 one, then the 10:14 one, then Epstein’s 6:19 AM forward and Osborne’s message beneath that. So the sheet nests the 10:14 message beneath the 9:55 one, and it places the 11:03 message below the 10:08 and 10:10 pair, not above it. Read at the image at 600 dpi.²³

The last message in the run is the one the desk anchored on. At a Sent: field of 8:35 PM on 3 January 2015, Wolff wrote:

I would not act till then. And I would hit very hard on media’s willing=ess to hide behind court documents and repeat all manner of preposterous a=legations. In that regard, I think you should prepare a detailed list=of the most preposterous aspects of all this—this will =ive the NYT and WSJ some pause.

¹ The = characters in these quotations are quoted-printable soft breaks carried onto the face of the sheets, and they are reproduced here as the sheets print them rather than repaired.

That document runs to four sheets, and the advice is not qualified, withdrawn or answered on any of them. Its second sheet opens with a redaction bar about two lines deep covering the attribution line at the head of the sheet, and carries beneath that bar the 10:14 AM message, then the 6:19 AM forward and Osborne’s original message; its third and fourth sheets, read at the image at 300 dpi, carry nothing but repeated copies of the confidentiality footer.²¹²⁴ So the advice message stands as the last thing in the thread as this document preserves it. What the document does not contain is any answer to it: no reply from Epstein, no statement, and no list of the kind the advice asks for, on any of its four sheets. Nor does one appear elsewhere in the release. "Michael Wolff" together with "draft statement", "issued a statement", "our statement", "the statement we", "here is the list" and "attached is the list" returns nothing on any of the six; and with the document’s own adjective for the list, run as "Michael Wolff" together with "preposterous", it returns two pages, this sheet and a scanned page of one of his own books.¹⁴

The attribution line barred at the top of the second sheet is not, in the event, a gap in the record: a separate production number carries the same 10:14 AM message as a document in its own right, with From: Michael Wolff printed in the clear and a Sent: field of 2:14 PM.²² This page takes the author from that sheet, not from the one that covers it.

February 2015: a magazine piece, a portrait, and “approvals”

The month after that exchange, the correspondence is about a piece Wolff was writing.

On 5 February 2015 Epstein wrote, at a quoted 6:40 AM: “can you send updated version.. need it in order to get approvals.” Wolff’s reply, timed Thu, 05 Feb 2015 12:30:58 +0000 and carrying an attachment the sheet names NYM--JE_1.rtf, does not answer the request. It reads, in full: “We are in Paris from 7th to 11th if you have any times…”²⁵

That sheet is the whole of what it shows. Epstein asked for an updated draft and gave a reason; the attachment travelled; the reply on the same sheet is about travel dates. The sheet does not record what the “approvals” were, who was to give them, or whether any were given, and it does not establish that Wolff put the piece to Epstein for clearance or gave him any control over it.

Five days later Wolff forwarded Epstein a thread about a photograph. Under the subject “Fwd: Jeffrey Epstein / portrait”, timed 10 February 2015, he wrote “Let me know what you want to do about this. Cold in Paris. m” above his own messages to a magazine’s photo desk: “I’ve sp=ken to him and he says he will arrange. Obviously can’t do more than t=at. But will try again”, and, four days earlier, “I have pinged him. Let me know what you hear.”²⁶ A separate sheet of 20 February 2015, read at the image, carries Wolff’s own message, quoted on it as sent at 1:18 PM beneath his one-word reply “Fab!”: “Jeffrey says you’re arranging details for this photo shoot with New York Magazine we’re planning…contact is…” and then the magazine contact’s name. The To: line of the covering message is barred, and the name and address on the attribution line of the message it answers are barred; this page identifies neither person.²⁷

The fact-checking correspondence that belongs to the same piece, the New York Magazine query list of March and May 2015 and the draft article found among Epstein’s files, is set out on Thorbjørn Jagland’s page, which cites it to sheets read at the image there. This page does not re-derive it.

On 13 February 2015, in the same weeks, Epstein sent Wolff a link to a Daily Caller piece and Wolff answered: “Roger Stone! Oy. And speaking of conspiracists!“²⁸

15–16 December 2015: “Heads up”

The third anchor is a three-message exchange over eighteen hours, and it is the clearest single instance on this page of advice recorded without an outcome.

At 8:00 PM Eastern on Tuesday 15 December 2015, at a Sent: field of 12/16/2015 1:00:36 AM on the House Oversight copy, which is one hour after midnight UTC, Wolff wrote to Epstein under the subject “Heads up”, marked “Importance: High”:

I hear CNN planning to ask Trump tonight about his relationship wit= you—either on air or in scrum afterwards.

²⁹³⁰ Epstein answered at 04:52:25 UTC on 16 December: “if we were able to craft an answer for him, what do =ou think it should be?”³⁰ Eleven and a half hours later, at a Sent: field of 4:27 PM on the EFTA copy read at the image and of 4:26:32 PM in the extraction of the House Oversight copy,⁹ Wolff replied:

I think you should let him hang himself. If he says he has=‘t been on the plane or to the house, then that gives you a valuable P= and political currency. You can hang him in a way that potentially genera=es a positive benefit for you, or, if it really looks like he could win, y=u could save him, generating a debt. Of course, it is possible that, when =sked, he’ll say Jeffrey is a great guy and has gotten a raw deal and i= a victim of political correctness, which is to be outlawed in a Trump reg=me.

⁸ That quotation is taken from the page image of the copy at EFTA02346409, which carries quoted-printable damage throughout; the extraction of the copy in the House Oversight production carries the same message without that transfer damage, but that sheet cannot be read at the image by anyone, because its endpoint returns HTTP 403, re-measured 11 September 2026. Every comparison between the two copies on this page is therefore an extraction set against an image, not two images.⁸⁹

The document records advice and nothing else. It does not establish that the question was put to anyone, that any answer was crafted, or that any use was made of the suggestion. The thread survives on five pages in all: the two House Oversight sheets, their two EFTA copies, and one further row the index holds under a name that is not a production number and that has no entry in the document table. Those five were found by searching for the thread’s own quoted timestamp strings rather than for phrases from its text, for the reason set out below, and nothing later in the thread stands on any of them.¹⁴

2016 and 2017: Washington, and offers to introduce

Through 2016 and 2017 the messages are mostly short, and they are mostly about access.

On 9 March 2017, replying to an Epstein message of 4 March that asked “if you get a chance see if you can gl=am the name of the next us ambassador to INDIA”, Wolff wrote:

Ok, seeing Bannon today, so I will float. Also, there’= a Washington PR type named Juleanna Glover with whom you might want to ch=t. Elon Musk basically owns her now and I’m not sure she can take on o=her clients, but she will have a valuable perspective on how you get to wh=re you want to be.

N=C. Was at WH all week. What are some of the names being discussed and I wi=l fish this week?

³¹ The sheet prints N=C. and WH, with the = a quoted-printable soft break; it does not expand either, and it does not say who was at WH. What the sheet establishes is that Wolff wrote that paragraph. It does not establish that he obtained anything, passed anything on, or had any standing anywhere; no answer to it appears on the sheet, and the searches below returned no later message reporting a result.

On 23 April 2017, under the subject “PR”, the whole of a one-sheet message reads:

Just had a chat with Hiltzig. You should definitely speak =o him. How would you like me to set it up?

³² The surname is printed “Hiltzig” on that sheet, read again at 600 dpi to settle the final letter, and the same spelling appears on a second sheet nineteen months later.³²³³ A proposal to set up a conversation is not a conversation, and no sheet read for this page records that this one took place. A Lesley Groff schedule circulated on 4 May 2017, read at the image, sets out two days under bold, italic, underlined headings reading “Thurs. May 4, 2017 NY” and “Fri. May 5, 2017 NY”. Under the Thursday it lists “12:30pm LUNCH w/Michael Wolff and Juleanna Glover”; under the Friday it lists “8:00am BREAKFAST w/Tom Pritzker”, “4:30pm Appt w/Matthew Hiltzik and his colleague, Ben Sosenko”, the surname spelled with a final k where the two Wolff sheets spell it with a g, and “6:30pm Doors Open for Terje’s Opening of: Oslo”. A contact value in parentheses is barred after several of the entries. On these schedules the time is set bold at the head of each entry, a capitalised meal word after it is bold too, and the remainder of the entry is regular italic; two entries on this schedule depart from that and are set wholly bold. The schedule runs to a second sheet, and that sheet says what “Oslo” is: “7:00pm Performance of Oslo”, “Lincoln Center Theater’s Broadway Production”, “At the Vivian Beaumont Theater at Lincoln Center”, the theatre’s street address, “YOU HAVE 4 SEATS” and “Jojo to drive you!!!” Nothing on that second sheet is covered by the producing party.³⁴ The Hiltzik appointment is on 5 May, twelve days after Wolff’s message. The lunch entry the day before is the one that closes a loop this page would otherwise leave open: Juleanna Glover is the person Wolff had named to Epstein on 9 March, quoted eight lines above, and eight weeks later Epstein’s office books a lunch with Wolff and her. A diary entry is a plan. The sheet establishes the booking; it does not establish that either appointment happened, or that the March message produced the May lunch.

The Friday evening entry is the one occasion on this page where the record carries the invitation, the diary and the acknowledgement. On 20 April 2017, at a quoted “On Apr 20, 2017, at 22:42”, Epstein wrote to Terje Rød-Larsen: “can we invite brad karp and michael wolfe to oslo”, the surname spelled with a final e on that sheet, read again at 600 dpi to settle the letter. Rød-Larsen answered the next morning: “They will be invited, but I need email addresses. There is now a full house, but I am holding 40 seats for last minute guests. I will send you the guest list as it looks now later today. It’s pretty impressive! I have reserved 5 seats for you.” Later that morning the addresses went back under the line “Hello Terje. Email addresses below:”, with Brad Karp and Michael Wolff named and both address values barred; the From: line of that message is barred and this page does not supply it.³⁵ Two weeks later the schedule lists the doors and the performance, and the morning after it Wolff wrote to Epstein: “That was a treat (and I usually hate the theater). And really liked Tom Pritzker. Are you leaving or staying for a bit?”, above Epstein’s message of the night before, “hope you guys had fun”.³⁶³⁷ The sheets establish an invitation asked for, granted, and acknowledged the next morning as a treat. They do not establish who else was there, and the breakfast with Tom Pritzker that morning and the theatre opening that evening are two separate entries on the schedule; the acknowledgement names both men and neither entry, and this page does not assign it to one.

Earlier that year, on 10 January 2017, Wolff had written: “Sorry, I was supposed to tell you that Ailes would like to organize something with you for the next time your down, he was in bad shape over his leg problems this weekend. / But look forward to telling you about the six-hour dinner I had with Ailes and Bannon last week at my house.” Epstein answered “ny just landed 430pm”; Wolff, “If you’ve got time this week, let me know. m”; Epstein, “today 430?”; Wolff, “great. see you then.”³⁸ Epstein then forwarded the exchange to a recipient whose address the production bars.³⁸

In October 2017 the brokering is visible end to end, and it ends in a refusal. As the forward preserves it, Wolff opened it. At a header of Date: Sun, 8 Oct 2017 09:42:39 -0400, on the second sheet of the forward, he wrote to Steve Bannon under the subject “How’s it looking…?” a message whose whole body is “lemme know”. Bannon answered “I’m in CT w/ donors to my revolution now all day…can we meet tom=rrow?”; Wolff replied “Sure. Give me a target window?”; Bannon asked “Did u see prince in nyt and djt tweets on corker???”; Wolff answered “Yes!” and then “JE leaves for Paris tomorrow at 9:30AM. Could meet before or tonight.” At 10:38 PM Bannon wrote back: “Can’t make tomorrow / Can we resked for when he returns”. The next afternoon Wolff forwarded the whole chain to Epstein.³⁹

That thread is worth reading beside every calendar entry on this page. It is the shape of the record: an arrangement proposed, discussed, and, in this one case because somebody said so in writing, not kept.

2017 and 2018: the book

From late 2017 the subject of much of the correspondence is Wolff’s own book.

On 29 November 2017 Wolff wrote about a broadcaster’s dismissal: “I think Matt will be bigger than H=rvey. Matt must have had a knot in his stomach when he had to report about=Harvey and Charlie, et al. Andy Lack must have had a knot in his stomach w=en he fired Matt. It’s extraordinary. And annoying: Matt had just agre=d to an exclusive interview with me about the book on publication day.”⁴⁰

On 4 January 2018, at a quoted 8:41 PM, Epstein asked: “am i mentioned in the book? .” At Fri, 05 Jan 2018 11:04:21 +0000 Wolff answered, and the whole message is one line:

Just passing reference that you, Tom Barrack, and Trump were once friends. One mention. No inference.

⁴¹ The same week Kathy Ruemmler sent Epstein a link to a trade-press interview with Wolff, writing: “Now being reported that T=ump’s lawyers sent a letter to MW’s publisher demanding th=t they not publish the book.”⁴²

On 24 January 2018 Lesley Groff wrote to Wolff: “Good morning Michael. Jeffrey is wondering if we could get 6copies of your book signed this morning for one of his assistants to take with her to Paris today. She could even bring over the copies to you and have you sign if possible? Please let me know! Thanks”. Wolff replied at 9:21 AM: “Yes. Can they be here by 11:00 (alternatively could do at 12:15)? And could you send me a list of who I should sign the books for?” Groff answered: “Fantastic! Where should ▮ meet you at 11am? Or alternatively at 2:15? Trying to get List of names. Will circle back! Thank you!” The name of the person who was to collect them is barred.⁴ The sheet records the arrangement. It does not record the collection, and no list of names appears on it.

On 2 February 2018 an introduction he asked for was made, and the record carries all three steps on three Sent: fields of the same class. At a Sent: field of 4:11 PM Wolff wrote to Epstein under the subject “A couple of things…”: “Thanks for Wed night, totally worth it! Can you put me in =ouch with Brad?”⁴³ Epstein’s introducing message survives as a document in its own right, addressed to “Brad S Karp; Michael Wolff” with no subject label on the sheet at all, at a Sent: field of 6:16 PM, and the whole of its body is “brad- michael , micahel brad”.⁴⁴ At a Sent: field of 6:24 PM Brad Karp answered both of them: “Thanks, Jeffrey. / Michael, Very much enjoyed meeting you and would love to catch up. / Brad”.⁴⁵ So the sequence is 4:11 PM → 6:16 PM → 6:24 PM: two hours and five minutes from the request to the introduction, and eight minutes from the introduction to the reply. All three are Sent: fields of the kind this page establishes as universal time, and the introducing sheet and the reply sheet each carry an Apple date-received integer beside the field. On the reply sheet the two agree to the minute. On the introducing sheet they do not: the integer is 1517595331, which is 18:15:31 UTC, against a field of 6:16 PM, so it falls twenty-nine seconds short of the minute the field prints and the two agree only to the hour.⁴⁴⁴⁵ The same introducing message is also quoted on the reply sheet, with a Date: line of “Friday, Feb 02, 2018, 1:15 PM”; that is a quoted header and is given no zone here, and the sequence above does not rest on it. The opening words of that request are themselves an acknowledgement, and they are tied to a diary entry. 2 February 2018 was a Friday; the Wednesday before it was 31 January, and the Epstein-office day sheet for Wed. Jan. 31, 2018 NY, circulated the previous afternoon, New York time, reads “7:00pm DINNER w/President Lajcak, Brad Karp and Michael Wolff”, with a contact value in parentheses barred after each of the three names.⁴⁶ So the message that asks to be put in touch with Brad thanks Epstein for the evening for which Epstein’s own office had booked a dinner with Brad Karp and Michael Wolff. What the day sheet establishes is the booking; what the message establishes is that Wolff wrote, two days later, that Wednesday night had happened and had been “totally worth it”. Neither sheet says who in fact sat down.

Four months later, on 6 June 2018, Wolff wrote to Epstein, in a message quoted in full on Epstein’s reply sheet:

Understand Brad is com=ng to your house at 10 on Friday with KR. Might you have a corner where Br=d and I can talk just after? I don’t want to be seen with him (I.e= not go to his office) but want to give him an idea what this article will=be about. Also, he doesn’t want KR to know anything about this.

Epstein answered “whatever you want . before after ok with me.” followed, after a quoted-printable artefact the sheet prints as =C2 and one unprintable glyph, by “anytime fine.” The initials in the request are not expanded on this page. No sheet read here records whether that further conversation happened, whether the article was described, or whether the third party the message names was told.⁴⁷

An introduction of a different kind is acknowledged four years earlier. On 8 February 2014 Wolff wrote to Epstein: “Woody and Soon-Yi came to dinner this week and seemed grateful for the piece I wrote about them. I’ve been reluctant though to do more—that is, not sure they want more. What do you think? Is the document something they want in circulation? I can do a column for Monday morning that incorporates it. And by the way, thanks for introducing us. They both are terrific. The injustice and illogic of all this is infuriating. Are you in Paris?” The sheet does not say when the introduction was made, and names no party to it beyond those the message itself names.⁴⁸

On 12 February 2018 a longer message shows him drafting strategy for a third party. Epstein had proposed “an independent investigation. funded by an amazon type. run by a female lawyer or former judgeetc. so woody can say after two months , they concludedd such and such…”. At 7:59 AM Wolff answered:

My concern is that this would be seen as just relitigating—a new investigation rehashing the same investigations that have already taken place. What’s more, it is unlikely the Farrow family would participate, so it’s illegitimate from the get-go. If it came out in his favor it would be judged as bias in his favor. On the other hand, this would be no worse than the current situation and might be a little better. But ultimately it doesn’t change the narrative: He ran off with his daughter and therefore he is capable and likely guilty of anything.

Epstein replied: “agreed, and amazon, told him to DO nothing. so he now has to follow their lead. sad”.⁴⁹ Several personal names inside the earlier message on that sheet are barred by the producing party, and none is supplied here. The subject line is two pairs of initials; this page does not expand them.

On 21 February 2018 Wolff forwarded Epstein, with no covering text of his own, a message he had received from Juleanna Glover asking “For some reason BBC is asking If I would put them in touch w you re Jeffrey. Do you want me to do this?”⁵⁰ The sheet does not record an answer.

A third introduction was made on 19 May 2018, in the same two-name form, and the third party credited Epstein by name. At Date: Sat, 19 May 2018 17:40:53 +0000, addressed to “Ken Starr” and “Michael Wolff” and with the Subject: label printing no value and no bar, Epstein wrote in full: “Ken - Michael , Michael/ Ken”.⁵¹ Wolff answered Starr, in a message the reply sheet quotes as sent “On May 19, 2018, at 1:14 PM”:

Ken, delighted to meet. Jeffrey has long told me about your friendship and the respect he has for you. If you send me your mailing address, it would be my pleasure to send you my recent book about the Trump White House. I’m now at work on a sequel focusing on the legal case against Trump and Trump’s response to it. Any counsel and background you might provide would aid me enormously and enrich the book. I, of course, would be willing to talk with you on an entirely off-the-record basis or under any arrangement that suits. Thanks in advance and I look forward to the chance to chat. All best, Michael

At Date: Sat, 19 May 2018 20:25:18 +0000, two hours and forty-four minutes after Epstein’s message and measured between two explicit +0000 headers both read at the image, Kenneth Starr replied to Wolff, copying Epstein: “Michael: / Excellent. With thanks to Jeffrey, I’m delighted to come into your orbit, and look forward to our conversation. / My mailing address is: / Warm regards, Ken”.⁵² An address value in that message is barred by the producing party and no value is taken from it.

Of the three introductions the record shows being made, this is the one in which the third party says what the introduction was worth to him, and it is the second in which Epstein’s introducing message is the same construction: “brad- michael , micahel brad” in February, “Ken - Michael , Michael/ Ken” in May. Both are two names, each given to the other, and nothing else. Twice in 2018, on 2 February and 19 May, Epstein made an introduction by sending both parties a message whose whole body was their two forenames, and on both occasions the third party answered both men within hours. That is what the sheets establish about the form. It is not a finding about how often the form worked, and no sheet read here records what either man had asked of Epstein beforehand beyond the February request quoted above.

And on this one the third party himself reports back to Epstein what came of it. On 13 June 2018 Starr forwarded to Epstein, under the subject “Fwd: Possible quote…”, a message Wolff had sent him that morning at a header of “June 13, 2018 at 10:18:43 AM EDT”, one of only three explicit local-zone labels on any sheet read here. Wolff’s message offers Starr a quotation for approval: “Don’t hesitate to say no if it’s at all uncomfortable”, then the proposed passage, “Might such an expansive claim win, yes. Should it win, no. But everyone knows the reality: weak president’s lose cases, strong president’s win them,” said Ken Starr, the former independent prosecutor who might know more about the legal pursuit of the president than anyone else on earth, and then: “And thanks so much for chatting the other day. I look forward, if you are game, to much more. m” Above it Starr wrote to Epstein: “Jeffrey: / FYI. I gave Michael the all clear. We’re going to continue to chat. / The “expansive claim” is the prosecutorial view that various presidential actions, eg firing Comey, may have constituted obstruction of justice. Bad theory. / Hugs, Ken”.⁵³

So on this one sheet the whole sequence closes: an introduction made by Epstein in a single line of two names, which no sheet read here records anyone asking him for; a conversation that had taken place by 13 June, in Wolff’s own words; a quotation put to the source and cleared, in the source’s own words; and the source reporting all of it back to the man who made the introduction. It is the only introduction on this page whose outcome is recorded by the third party rather than by Wolff. The February 2014 introduction is reported by Wolff himself, “Woody and Soon-Yi came to dinner this week”, and on the February 2018 one no sheet read here records anything past Karp’s reply. This one runs through Epstein at both ends.

Two later sheets bear on how far it went in person. On 9 July 2018, in a thread otherwise about a Supreme Court nomination and about Steve Bannon, whom the 9 July sheet names only as “SB” and whom the next day’s sheet in the same thread names in the clear, Wolff wrote “If I’m still here would love to meet Starr”, and Epstein answered “of course”.⁵⁴ The next day Wolff wrote: “I’m going. A full Bannon schedule tomorrow. Give my best to Mr. Starr. I will catch up with him soon.”⁵⁵ Those two messages are the last word on it in the sheets read here, and they leave the question of a meeting open: on Wolff’s own words of 9 July 2018 the two had not met in person, and no sheet read here records that they did afterwards.

Epstein had named Starr to Wolff the previous November. On 26 November 2017 Epstein wrote to Wolff: “I spent many days now with Ken Starr, you might want to suggest to SB that it is crazy that no one on his side has spoken to the person who knows more about independent counsel investigations than anyone else by far. boundaries. tactics. procedures.”⁵⁶ Whether that “SB” is the “SB” of the July 2018 thread is not settled by any sheet read here, and the initials in this message are not expanded on this page.

The last year in the record: press strategy before the arrest

Through 2018 and the first half of 2019 the correspondence returns to the thing it began with.

On 24 August 2018, after Epstein wrote “State investigation s right on time—”, Wolff answered: “So many shoes to drop. Let me know if you have time for a =ea this afternoon.”⁵⁷

On 2 December 2018, in a thread whose subject is “thoughts”, Epstein sent a long message setting out points he thought reporters covering his case ought to make, among them that “the idea that mr epstein was an informant. has absolutely no foundation in fact whatsoever.” At 9:02 PM Wolff replied:

You’re thinking of the investigative journalist as having a higher purpose instead of being a player in a greater drama. Your enemies have supplied them with value. So the issue is what you can offer them. Let’s chat tomorrow.

Five minutes later Kathy Ruemmler answered both of them, copying Darren Indyke: “J, / Please have ▮ put a call/meeting together tomorrow so we can discuss collectively. Michael is correct that this should not be done piecemeal. / K.” The name of the person who was to arrange the call is barred.⁵⁸ Epstein forwarded the chain onward the same night.⁵⁸

Two days later, on 4 December 2018, Epstein asked “do you think the press would react to the fact that all the settlement money is going to the attorney and none to the girls.? react to the fact that the case was settled but the lawyers wanted to trumpet their success.” Wolff answered at 8:25 AM:

I think it’s a useful point, potentially a powerful one. But I don’t think anything is going to get attention now. I would look for some reporter to do a more nuanced post-morten on the case—with it’s Trump overtones, legal joustings, metoo-isms, and profit-motives. WSJ is probably right place.

Epstein asked “should i , or you reach out to matt. or sean . or ?”, and Wolff replied: “I think you should reach out to Matt. Moonves’ guy, who I had lunch with yesterday, says he would probably use Hiltzig.”³³ The sheet gives the forenames as it gives them and this page supplies no surnames. Nothing on the sheet records that anyone was approached.

On 1 February 2019 Epstein forwarded Wolff, under the line “have fun”, a long message he had sent to his own address at 5:18 AM the same morning attacking the investigation into him; it is the message from which the phrase “operation leap year” is taken on the page about that investigation, and one personal name in it is barred.⁵⁹ The sheet carries no reply.

On 27 February 2019, under the subject “script draft”, Wolff was working on a video for a third party. At 8:07 AM he wrote “This is done very quickly. Don’t have the facts here. Trying to get a tone and strategy that you can start to discuss. What did Woody say?”; Epstein answered “woody thought it should be in an interview format with you as the interviewer , his views”; at 8:51 Wolff replied:

I’d disagree. If you have an interviewer you have to submit to a tough interview, otherwise it really looks canned. But, more importantly, I think Woody doesn’t understand the social media thing. It isn’t a substitute television. It’s power is really from one person looking into the camera and talking, computer to computer (or phone to phone), directly to another person. You don’t want to disintermediate. It has to look like you are really there, all in, and exposed.

Epstein: “i agree with you , his take was to tell eveyone how i ve been mistreated. if that gives you an idea of his strategy :)”. Wolff: “Yeah, that needle’s gonna require some finer thread.”⁶⁰ The sheet records a draft being discussed. It does not record a video.

In April 2019 the exchange about Epstein’s response to the Miami Herald reporting, in which Wolff sent a long comment on a draft and copied Bannon, is set out on Steve Bannon’s page and on Julie K. Brown’s, each cited there to a sheet read at the image. This page does not re-derive it.

The last sequence read for this page is from 27 and 28 May 2019, and it is about a book again. It opens on the forward’s second sheet, with Lesley Groff writing “Hello Michael…Jeffrey would like us to pick up a copy of your book…when would be convenient to send someone your way? Please advise an address to pick up from!” over the signature “Thanks, / Lesley / Assistant to Jeffrey Epstein”. Groff then wrote “HI Michael…checking in…Jeffrey is itching to get his hands on your book! Any word?”; Wolff answered “Should have one by 3:30. Should I let you know then?”; and at 2:13:14 PM EDT, one of the few quoted headers in this correspondence that carries an explicit zone label, he wrote: “3:30 is good. Who is coming for it? Need high security!” Groff then instructed a colleague: “Jojo will pick up book now and bring to you. Bella please scan and email JE forst 40 pages them fed ex for morning delivery tomorrow to PB and provide me the tracking number. Thx!”⁶¹

What the day sheets record, and what they do not

Wolff’s name runs through Epstein’s diary for most of a decade. A diary entry is a plan. This page does not treat any of them as evidence that a meeting took place, and the reason is on the sheets themselves.

The schedule for Thursday 9 January 2014, circulated by Epstein’s office on 7 January, sets out the day in italic type and opens: “9:00am Breakfast w/Michael Wolff (▮)”, with a contact value in parentheses barred, followed by six further appointments on that sheet and three more on a second sheet.⁶ It records what the office intended. It does not record what happened, and no message read for this page reports back on that morning.

A dinner list of 22 January 2014 makes the same point in the sheet’s own words. Under the heading “Guest List at the Moment:” it names fifteen people, Wolff among them, and annotates two of them “(but maybe not? I can’t tell by JE email)” and “(still need him to confirm)”.⁵ The list is explicitly provisional on its face.

Against that, the release does record occasions as having happened, in his own words. The searches set out below returned seven short messages from Wolff that are first-person acknowledgements rather than plans, and all seven were read at the image for this page. Five of them are what the twenty-six-form search returned; the sixth is the May 2013 message this page already carried, which none of those twenty-six forms reaches; and the seventh was returned by "was a treat", one of twenty further forms run afterwards and reported below. This census is a census of what those tests returned, not a statement about the release. The count on this page has now been one, then six, then seven, and each rise came from a test the previous pass had not thought to run:

• 2 September 2011, subject “thanks”, the whole body: “Well that was great fun. I loved that guy. And Victoria loved you. Soon. m”. This is the one the other party answers on the same sheet: Epstein replied, and the whole of his reply is “i am thrilled that you looked so happy”.⁶²
• 16 February 2012, subject “Brooklyn”, the whole body: “That was fun”. Epstein answered “Agreed”.⁶³
• 3 May 2013, subject “thanks…”, the whole body continuing the subject line: “…for the hot dog. I will be in Paris with Victoria from June 12-17. Best, m”.²
• 27 January 2014, at Date: Mon, 27 Jan 2014 14:15:56 +0000, the whole body: “Super! Really nice evening. Thanks. When are you back?” It is the earliest message in the chain that preserves it; the exchange arranging a visit for four o’clock follows it rather than precedes it, so the acknowledgement looks back and not forward.⁶⁴
• 6 May 2017, at Date: Sat, 06 May 2017 13:40:25 +0000, the whole body: “That was a treat (and I usually hate the theater). And really liked Tom Pritzker. Are you leaving or staying for a bit?” Epstein’s message the night before, which the sheet quotes as “On Fri, May 5, 2017 at 11:38 PM” and which also survives as a document of its own at Date: Sat, 06 May 2017 03:38:04 +0000, is four words: “hope you guys had fun”.³⁶³⁷
• 2 February 2018, subject “A couple of things…”: “Thanks for Wed night, totally worth it!”⁴³
• 6 March 2018, Wolff’s own subject line “That was fun”, the whole body: “Love those guys. Also…having trouble getting a response =rom Simpson. Any ideas?”⁶⁵

Two of the seven are tied to a diary entry printed or quoted on this page, and five are not.

• The 6 May 2017 message is the clearest. The Epstein-office schedule for Fri. May 5, 2017 NY, circulated the previous morning, New York time, and read at the image for this page, lists “8:00am BREAKFAST w/Tom Pritzker” and “6:30pm Doors Open for Terje’s Opening of: Oslo”, and its second sheet gives the 7:00pm performance, the producing theatre company and the house.³⁴ The next morning Wolff names the theater and Tom Pritzker, and Epstein had written the night before hoping they had had fun. What the two sheets together establish is that an occasion listed in the diary is acknowledged afterwards as having happened; the message does not say which of the day’s entries he means, and this page does not assign the two things he names to one entry rather than the other.
• The 2 February 2018 message opens “Thanks for Wed night, totally worth it!” The Wednesday before that Friday was 31 January 2018, and the Epstein-office day sheet for that day lists “7:00pm DINNER w/President Lajcak, Brad Karp and Michael Wolff”.⁴⁶ The same message then asks to be put in touch with Brad.

A third is adjacent in date but not tied on its face. The dinner guest list of 22 January 2014 is for “Dinner on Sun 26th” and names Wolff fourteenth;⁵ the “Really nice evening” message is timed Mon, 27 Jan 2014 14:15:56 +0000, the following morning. But that message names no occasion, no place and no person, and the list is provisional on its own face, so this page does not treat the two as one event. The remaining four, September 2011, February 2012, May 2013 and March 2018, fall on no date for which any diary entry on this page exists.

None of the seven sheets says where, or in whose company. What they establish is that the release is not silent about whether occasions happened: it records them after the fact, in the first person, on the seven sheets above, and on two of those occasions a diary entry for the same day or the day before is in the record as well.

A further scheduling reference sits on a page already published: a Groff daily schedule of 10 December 2012 whose subject line is his name, on Harry Beller’s page. The dinner entry of 31 January 2018, quoted in full two sections above from the sheet itself, is also carried on Miroslav Lajčák’s page. Both are diary entries and are described as such in both places.

Advice and outcome: what this page does not establish

Set out together, because it is the point of the page:

• A recommendation is not a result. Of the messages quoted above, the press-response advice of January 2015, the debate-question advice of December 2015, the “let’s chat tomorrow” of December 2018 and the interview-format advice of February 2019 are all recorded on sheets that carry no answer and no later report. The searches set out below were built from what an answering document would say rather than from the words already on this page, and on those searches no answer returns.
• A plan discussed for two years is not a plan carried out. The reputation project is the one strand where the release does hold later documents, and they run the other way: in September 2013 the outline Wolff had promised in June 2011 had still not been sent.¹⁰¹²
• A draft request is not editorial control. The sheet of 5 February 2015 shows Epstein asking for an updated version “in order to get approvals” and shows Wolff’s reply not addressing the request.²⁵ It does not show what the approvals were, who gave them, or whether the piece was changed.
• A proposal to set up a conversation is not the conversation, but this page does not say the record is silent about outcomes, because on introductions it is not. “How would you like me to set it up?” is the whole of the April 2017 message about Hiltzig, and no sheet read here records that that conversation took place; a calendar entry twelve days later is a calendar entry.³²³⁴ Against that, three introductions are recorded as made: on 2 February 2018 Wolff asked to be put in touch with Brad Karp, Epstein wrote the introducing message at a Sent: field of 6:16 PM, and Karp answered both men eight minutes later, “Very much enjoyed meeting you”;⁴³⁴⁴⁴⁵ on 19 May 2018 Epstein sent the same two-name message to Kenneth Starr and Wolff, and Starr answered both, “With thanks to Jeffrey, I’m delighted to come into your orbit”;⁵¹⁵² and on 8 February 2014 Wolff wrote, “by the way, thanks for introducing us”.⁴⁸ And on one of the three the record carries the outcome as well: by 13 June 2018 Wolff and Starr had spoken, Wolff had put a quotation to Starr for approval, and Starr had written to Epstein, “I gave Michael the all clear.”⁵³ The distinction this page keeps is between a proposal and its outcome, not a claim that outcomes go unrecorded, and where the record carries an outcome, this page carries it.
• Initials and bare surnames in this correspondence are not identifications. The sheets carry “tina”, “Matt”, “sean”, “Moonves’ guy”, “Gates”, “WA and SY”, “Hiltzig”, “N=C”, “KR” and “SB”; this page prints each as the sheet prints it and expands none of them but one. The exception is the “SB” of the July 2018 thread about meeting Kenneth Starr, where the next day’s sheet in the same thread prints “Bannon” in the clear: that expansion rests on a document and is stated where it is made.⁵⁴⁵⁵
• A message asking what names are being discussed is a message asking. The March 2017 sheet establishes that Wolff wrote “What are some of the names being discussed and I wi=l fish this week?” It establishes nothing about what he had, or got.³¹

What the search returned

The corpus of released files was searched with the FTS5 index on 10 September 2026 in the following forms. Counts are given as pages and as distinct documents, over the whole index, which covers the House Oversight and DOJ-OGR productions as well as EFTA.

• "Michael Wolff": 5,620 pages in 5,388 documents
• Wolff (token): 5,833 pages in 5,529 documents
• Wolff* (prefix): 5,878 pages in 5,556 documents
• "Wolff, Michael": 103 pages in 97 documents
• "M. Wolff": 0; "Mike Wolff": 0
• "Fire and Fury": 130 pages in 127 documents; "Burn Rate": 56 pages in 51 documents; "The Man Who Owns the News": 5 pages in 4 documents; Siege together with Wolff: 63 pages in 49 documents. Scanned pages of his books account for a substantial part of the raw count and are not records of anything about Epstein.¹⁴

Damaged spellings. python3 scripts/ocr_variants.py "Michael Wolff" --wide --count generated 37 forms and returned eleven non-zero; the other 26 returned zero pages against the full index. Each non-zero form was sampled at the rows before any decision was taken about it:

form	pages	decision
Michael Wolf (one f)	196	rejected — see below
Michael Woiff	23	accepted: Epstein correspondence and office scheduling
ichael Wolff	15	accepted: quoted headers inside Epstein threads
Michael olff	12	accepted: Epstein correspondence
Michae Wolff	11	accepted: quoted headers inside Epstein threads
Michaei Wolff	7	accepted: day sheets
Michacl Wolff	3	accepted: day sheets and one Epstein thread
Michael Wolfff	3	accepted: day sheets and one Epstein thread
Michael Wo1ff	2	accepted: two Epstein threads
Michel Wolff	2	accepted: one Epstein thread, one press clipping
Miehael Wolff	1	accepted: one day sheet

The union of the canonical name with the ten accepted forms is 5,648 pages in 5,413 documents: 25 documents, 0.5%, beyond the canonical population. That is at the low end of the 0–4% band this desk has measured for damaged-spelling loss, and it is well below the 2.7% an earlier text-layer pass recorded, because that pass included the one-f form and this one does not.¹⁴

Michael Wolf with one f is rejected, and the reason is not OCR. Its 196 pages include 125 that carry no two-f spelling anywhere on the page. Those 125 are spread almost evenly across the productions that carry this subject: 45 in the ninth data set, 36 in the tenth, 38 in the eleventh and 6 in House Oversight, which on one basis throughout (damaged-only pages over that production’s canonical pages plus its damaged-only pages) is 1.90%, 3.32%, 1.95% and 1.78%. A scan-batch OCR fault concentrates in one production; this does not. What the sample found instead is a second, different person of that name printed in the clear: an alphabetised roster of names and biographies, held in the release at several production numbers, carries the entry “Michael Wolf – CEO, Activate. fmr President and COO, MTV Networks. Fmr Prtnr, McKinsey. Former Senior Partner, Booz Allen & Hamilton”, set between two neighbouring surnames in clean typeset text with one f.⁶⁶ Twelve of the 125 pages carry that roster entry. They were enumerated by reading all 125 rows and testing five independent markers from the roster sheet: Activate, MTV Networks, Wolpe, Weymouth and Booz Allen. Thirteen rows match at least one marker and twelve match three or more; the thirteenth matches only the single token Activate, which appears on it as a company name in a separate column of a long alphabetical list, and it carries no roster entry.⁶⁷ A one-marker match is why this count read thirteen before it was enumerated. The rest of the 125 include a professional-network contacts export naming a Michael Wolf with a different occupation, and two long alphabetical name lists, all three read in the text layer only.⁶⁷

The form does also reach this subject: three messages that carry it are the same messages that survive elsewhere spelled with two fs: a March 2016 memo about a publisher, a February 2019 forward, and a March 2015 forward of a magazine’s fact-checking questions.⁶⁸ But a form that reaches two different people cannot be unioned into a count, and the rows it returns cannot be separated by the string alone. Day-sheet entries reading “LUNCH w/Michael Wolf” are therefore not attributed on this page to either man.

Homographs. The Wolff token returns 5,529 documents, of which 128 carry neither the canonical name nor any accepted damaged form. Ten families were named and each was confirmed by a separate query to lie wholly inside that residual: Guntram Wolff 1 document, Nathan Wolff 3, Lisa Wolff 1, Scott Wolff 2, Wendy Wolff 3, Steven D Wolff 4, Philip J Wolff 3, Kleinberg Kaplan Wolff 6 (a firm), Jocelyn Wolff 1 and Climenhaga Wolff 1, 25 documents in all, with four further documents using the possessive Wolff's. That leaves 99 documents in the residual which this pass did not attribute to any family, and they are not claimed as either this subject or anyone else.¹⁴

Act against mention. Sixty pages were drawn from the 5,620-page canonical population at evenly spaced ranks after sorting all matching row identifiers, and every one was read and classified by hand:

class	pages
a message Wolff himself wrote, reproduced on the page	34
addressed to him, with no passage of his on the page	6
a day sheet, schedule or diary entry	9
a guest or invitation list	1
a press clipping, book excerpt or book inventory	5
other people scheduling or discussing him	5
total	60

So 34 of 60 sampled pages, or 57%, carry an attributable message of his. An earlier text-layer pass, sampling 40 pages from the same population at different intervals, measured 22 of 40, or 55%; the two samples overlap only partly and agree closely. The nine calendar pages in this sample are not counted as acts, and several of them record an appointment as fixed rather than proposed. A calendar in a text layer is not proof that a meeting occurred, and promoting all nine would have raised the figure by fifteen points on a basis the documents do not support.

Absence, and what would disconfirm it. Three statements above are negative. An absence test built out of the words already on a page can only confirm the page, so for each of the three the question asked first was what a disconfirming document would say, in its own words; the search forms were written down from that answer before any of them was run, and every form is reported with what it returned whether or not it falsified the statement. The second test did falsify the earlier reading of this page, and the section above is the result.

1. Whether the reputation project was carried out. The disconfirming document would be an outline, a design, a live page, an invoice or a contract. "Michael Wolff" together with outline returns nineteen pages, and reading all nineteen resolves them into five exchanges: eight copies of the October 2011 thread, four of the June 2011 promise, two of the November 2011 costing, two of the August 2013 memo and three of the September 2013 chase. All nineteen are about an outline; none of them is one, and the last of them postdates the first by twenty-seven months.

2. Whether a scheduled meeting is recorded as having taken place. The disconfirming document would be a short message afterwards, first person and past tense, saying that something happened. Twenty-six forms of that shape were written down and then run, all of them derived from that description of a disconfirming document and none of them taken from a phrase this page quoted before the test was run. (Two of the twenty-six, "that was fun" and "nice evening", now appear in the page’s own prose, but they appear there because this test returned them; a later reader finding those strings above has found the test’s output, not its source.) The forms and their returns: "Michael Wolff" together with "that was fun" returns 14 pages, "nice evening" 21, "was great fun" 4, "see you soon" 4, "great seeing you" 2, "thanks for coming" 2, "enjoyed meeting" 1 and "totally worth it" 1; and the other eighteen return nothing: "so much fun", "what fun", "great evening", "lovely evening", "wonderful evening", "last night was", "thanks for last night", "thanks for dinner", "thanks for breakfast", "thanks for having me", "great to meet", "nice to meet", "really enjoyed", "thanks for the evening", "was terrific", "good seeing you", "nice seeing you" and "lovely to see you". Every returned page was read. The 14 are three threads: nine copies of a September 2014 message by a third party, three of Wolff’s March 2018 exchange and two of his February 2012 message. The 21 are twenty-one copies of one message of 27 January 2014. The 4 under "was great fun" are four copies of the 2 September 2011 message. "totally worth it" and "enjoyed meeting" are the two February 2018 sheets. The remaining eight are other people’s words or book text: "see you soon" is Wolff proposing rather than acknowledging, "great seeing you" is two copies of a message of 6 March 2018 to Wolff from a third party, and "thanks for coming" is a scanned book. Those returns are five of the seven acknowledgements set out above, and they are the whole of what these twenty-six forms found in his own words.²¹⁴

   Twenty further forms were written down and run afterwards, on 11 September 2026, and they returned a seventh acknowledgement. They were derived the same way, from the same description of a disconfirming document, and written before any was run: "thanks again", "that was great", "was a treat", "wonderful time", "lovely time", "good time last", "enjoyed last night", "such fun", "terrific evening", "nice to see you", "great to have seen", "thanks for the lunch", "thanks for the drink", "fun last night", "loved meeting", "delighted to meet", "pleasure meeting", "thanks for everything", "what a night" and "great night", each in conjunction with "Michael Wolff". Sixteen return nothing. Four return something, and every returned page was read: "thanks again" 12 pages, of which ten are copies of a December 2015 to March 2016 enquiry Wolff made of a publisher’s publicity staff about another author’s book, and two are an Epstein-office scheduling exchange of 5 April 2014 in which the words are not his; "that was great" 6, which are the six copies of the 2 September 2011 message and are the reason the copy count in ⁶² is five and not three; "was a treat" 2, which are the two copies of the 6 May 2017 acknowledgement, the seventh, and a message no form in the twenty-six reaches; and "delighted to meet" 5, which are the Starr introduction of 19 May 2018 and are not an acknowledgement of an occasion at all. Twenty-six forms found five, twenty more found a sixth class of message and one further acknowledgement, and the question of whether the set is now complete is not one this page can close. That is why the census above is scoped to the tests and not to the release.³⁶⁵²¹⁴

   Two things this test exposed about the method, and they generalise. The copy of the 2 September 2011 message that this page actually cites is not among the four that "was great fun" returns, because its extraction reads “great lint” where the sheet prints “great fun”: a phrase search counts only the copies whose extraction survived. And the 27 January 2014 message is returned 21 times by the conjunction but 22 times by the phrase "Really nice evening" alone, because on one of those documents the body stands on a second sheet that does not carry the name. A two-term search is scoped to a page, and a message that straddles a sheet break can fall out of it.⁶⁴⁶²

3. Whether the December 2015 suggestion was used. Two disconfirming documents are possible: a later message in that thread, or a message afterwards reporting that the question was put or an answer used. Neither test is drawn from this page’s prose. The thread was identified by its own quoted timestamp strings, which belong to the documents rather than to this page and are near-unique across the release: "Dec 15, 2015 at 8:00 PM" returns 4 pages, "Dec 15, 2015 at 11:52 PM" 3, "12/16/2015 1:00:36" 1 and "12/16/2015 4:26:32" 2, and their union is exactly the five pages named above, with nothing later in the thread on any of them. For a report back, "Michael Wolff" was run with each of sixteen forms of what such a message would say: "he was asked", "they asked him", "asked him about", "he denied", "denied knowing", "never been on the plane", "town hall", "the debate last night", "what did he say", "did he say anything", "he said nothing", "nothing came of", "the answer we", "used the answer", "asked in the scrum" and "after the debate". Five returned anything, nine distinct pages between them, and every one was read: seven are scanned pages of published books or long-form press text, and two are an exchange of January 2018 about a television appearance. On these searches nothing in the release records that the question was put, or that any answer was used.¹⁴

How the reading for this page was chosen, and what it does not cover. Fifty-six documents were read at the image, over seventy-nine sheets, in three passes and a fix pass. Thirty-four documents over thirty-five sheets were read for the first draft, and that draft’s own account of how they were chosen is: two anchors the desk named, the third anchor sitting in a production whose endpoint returns HTTP 403 and not renderable; seven further copies of the same three threads, which is how the barred attribution on one sheet was resolved from another and how the December 2015 message became readable at an image at all; seventeen out of the sixty-page systematic sample; and eight reached by searching a strand the sample or an absence test had exposed. Two, seven, seventeen and eight make thirty-four. That split between the sample and the strand searches is the first draft’s own record and was not re-derived.

Eleven further documents, over fourteen further sheets, were read at the image in the second pass on 11 September 2026, and they are, individually: EFTA01854645, EFTA02550537, EFTA00684779, EFTA02533377 and EFTA02511806, which are five of the seven acknowledgement sheets; EFTA02533200, the February 2018 reply; EFTA01933332, the chain that carries the 27 January 2014 message in its place; EFTA00968035, the August 2013 sibling carrying “my pleasure”; EFTA00450519, the May 2017 schedule; EFTA02085392, the February 2015 photo-shoot sheet; and EFTA02646476, the June 2018 request. The last two of that list and the schedule had been cited unverified before. The three further sheets are the third and fourth sheets of the January 2015 anchor and a second sheet of EFTA01933332; the three copies of the 10:14 AM message were also re-read at 600 dpi, on sheets already counted. The May 2013 hot dog sheet is not among the eleven: it was read for the first draft.

Eleven further documents, over sixteen further sheets, were read at the image in the third pass on 11 September 2026: EFTA00696039 and EFTA01047230, the 6 May 2017 acknowledgement and the message it answers; EFTA02210485, the April 2017 invitation to the theatre opening, read again at 600 dpi; EFTA02532742, Epstein’s February 2018 introducing message as a document of its own; EFTA01055521 and EFTA01055612, the May 2018 introduction and Kenneth Starr’s reply; EFTA02629642, the June 2018 quotation clearance; EFTA01003826 and EFTA01005209, the July 2018 messages about meeting Starr; EFTA00946297, the November 2017 message that first names Starr to Wolff; and EFTA00468653, the day sheet for 31 January 2018. Five of the sixteen sheets belong to documents already counted, and all five are second sheets that no earlier pass had opened: EFTA00450519 p.2, which carries the performance entry quoted above; EFTA01939225 p.2, which carries three further appointments; EFTA02589079 p.2, which carries the earliest message of the October 2017 chain; EFTA00496165 p.2, which carries the earliest message of the May 2019 chain; and EFTA00968035 p.2, which carries the last three lines of the August 2013 memo. They were found by checking documents.total_pages against the page’s own pins. In the third pass that check was carried only over the documents that pass itself added; it has since been run over every identifier cited here, and what it found is set out in the note below. Thirty-four, eleven and eleven make fifty-six; thirty-five, fourteen and sixteen make sixty-five, and the fourteen further sheets opened in the fix pass make seventy-nine. Five thousand three hundred and eighty-eight documents is far too many to read, and this page’s reading is a floor rather than a census. What it is likely to have missed is anything in the great bulk of duplicate scheduling traffic, and anything whose only copy spells the name in a way none of the forms above reaches.

Note on the documents

Seventy-nine sheets across fifty-six documents were read at the image for this page, rendered locally, at 400 dpi except where stated. Ten sheets were read at 600 dpi. Six of them: the April 2017 “PR” sheet, to settle the final letter of a surname; the 21 April 2017 sheet, to settle the final letter of a surname spelled differently again; the June 2018 sheet, where the question was whether one stroke is a capital I or a lower-case l and the image does not settle it; and the three copies of the 10:14 AM message of 3 January 2015, where the finding is which soft break falls where, where the order of the nesting was read, and where the depth of one sheet’s redaction bar was measured.^323547212223 The other four were read at 600 dpi in the fix pass, to test the four places where a footnote asserts that a sheet prints a damaged form that is not a quoted-printable soft break: “I’m leave” and “St. Petersberg”, “didn;t” and “something to today”, “post-morten”, and “forst 40 pages”. All four are on the sheets, and so are “Reminer” and “Jefffrey” on the schedules; the one claim of that class that failed is the subject of item 14 above.^10123361 Fourteen sheets were read at 300 dpi: the third and fourth of the January 2015 anchor, where the only question was whether any message stands on them, and twelve of the further sheets described in the next paragraph.²⁴ The quotations above are taken from those page images. Eighteen footnotes below carry an UNVERIFIED marking or say in terms that a sheet was read in the text layer only, and sixteen of them cite at least one unrendered sheet by URL. Nine cite an unrendered sheet as their own source, and each says so in its own words and says what verifying it would cost: ¹⁶, ²⁶, ²⁸, ²⁹, ⁹, ⁴⁰, ⁴², ⁶⁷ and ⁶⁸. Six more carry an UNVERIFIED marking only as to further copies of a message whose own sheet was read at the image, and each names those copies by URL: ⁶², ⁵¹, ⁵², ⁵⁶, ³⁶ and ⁶⁵. Two carry no UNVERIFIED marking but say in terms that further copies were read in the text layer only: ⁶³, which names its copy by URL, and ⁶⁴, which names none. The eighteenth is ⁵⁹, which is UNVERIFIED as to the second sheet of the very document its first sheet was read from and gives that sheet no separate URL; it belongs to none of the three categories above, and an earlier version of this sentence left it out of all of them and counted seventeen. Counted by the UNVERIFIED label the figure is sixteen; counted by “the footnote’s own source sheet was not rendered” it is nine; counted by “cites a sheet that was not rendered by URL” it is sixteen again, and the two sixteens are not the same sixteen: because ⁵⁹ carries the label and names no unrendered sheet while ⁶³ names one and carries no label. Two of those are the House Oversight copies of the December 2015 thread, whose PDF endpoint returns HTTP 403; for both of them a renderable EFTA copy carrying the same message was opened instead, and the quotations are taken from the EFTA image.²⁹⁹ Both of those two have since been read at the image directly, on 2026-09-14, from the production’s own page-image store rather than through the /pdf/ route; see ²⁹ and ⁹. Every count in this paragraph is as of the pass that produced it and has not been re-derived; the standing correction is that a 403 on /pdf/ is a fact about that route and never a statement that a sheet cannot be read.

Second sheets, and the documents.total_pages check. The check was re-run in the fix pass over all eighty-five identifiers cited on this page. Twenty-one are pinned at a sheet short of their run; seven of those are cited here as unrendered in the first place, so no second-sheet question arises for them. That leaves fourteen image-read documents pinned short of their runs. One is EFTA01028632, already declared at ⁵⁹, whose second sheet is still read in the text layer only and is marked UNVERIFIED. The further sheets of the other thirteen, fourteen sheets in all, were rendered and read for this pass, and none of them carries a message: EFTA00468653 p.2, an Apollo Global Management confidentiality footer; EFTA02691062 p.2, the closing tag of an Apple property list; and EFTA00719949 p.2, EFTA00888548 p.2, EFTA00968026 p.2, EFTA01003826 p.2, EFTA01014613 p.2, EFTA01032286 p.2, EFTA01790536 p.2, EFTA01791592 pp.2-3, EFTA01933332 p.3, EFTA02199945 p.2 and EFTA02353297 p.2, all Epstein-house confidentiality boilerplate. The first two of those were read at 400 dpi and the rest at 300, which answers the only question asked of them. Two of the thirteen, EFTA00719949 and EFTA00968026, already described their second sheets in their own footnotes without having rendered them; those descriptions have now been read at the image and they hold. So the statement above that the check was run is now as wide as the check. The one further sheet still unopened is ⁵⁹‘s second, which that footnote marks UNVERIFIED and describes from the corpus text layer alone.

Time zones. The unlabelled Sent: field in these exports is coordinated universal time, and this page relies on that. Seven sheets read at the image print an Apple date-received integer beside the field, and six of them are relied on here: 20 July 2011 5:48 PM against 1311184087; 31 January 2015 9:14 PM against 1422738859; 16 December 2015 4:52 AM against 1450241545; 2 February 2018 6:16 PM against 1517595331; 2 February 2018 6:24 PM against 1517595852; and 7 June 2018 12:02 AM against 1528329737. On five of the six the integer and the field agree in UTC to the minute. On the 6:16 PM sheet they do not: 1517595331 is 18:15:31 UTC, twenty-nine seconds short of the minute the field prints, so there the two agree to the hour and no further. That is closer than the ninety-second gap on the seventh sheet, the February 2012 message at ⁶³, which this page declines to rely on for the same reason; it is still not equality, though the four-hour question these comparisons exist to settle is not affected by twenty-nine seconds either way.¹³⁶³ That seven is a measured figure, not an impression: the corpus row of every sheet pinned on this page was searched for the string date-received, which returns thirteen sheets, and six of those thirteen are copies this page cites as unrendered. A separate check runs the other way: the anchor sheet of 5 October 2011 gives Sent Wed 10/5/2011 4:19:44 AM, and a sibling copy of the same message, also read at the image, carries the header Date: Wed, 05 Oct 2011 04:19:44 +0000.⁷¹⁸ The times quoted inside the Gmail chains are a different matter and no zone is asserted for them. On the two January 2015 messages where the same message survives both as a Sent: field and as a quoted line, the quoted line runs about four hours behind the UTC field (exactly four hours on the 10:14 AM message against its 2:14 PM field, and four hours and one minute on the 11:38 AM message against the 3:39 PM field of EFTA01791592), which is not New York’s January offset, so each is given above as its own sheet prints it.²²²³ Three quoted headers carry an explicit local-zone label: EDT on a 2019 sheet and on the June 2018 sheet at ⁵³, and EST on the 31 January 2018 day sheet at ⁴⁶, which reads “Date: January 30, 2018 at 1:07:31 PM EST”. One further quoted header, on the second sheet of the October 2017 forward, carries the numeric offset -0400. Those four are the only zoned local times on any sheet read here. That bound was measured, not assumed: the corpus row of every sheet pinned on this page was swept for EDT, EST, PDT, PST, CDT, CST, MDT, MST, GMT, BST, CET and CEST and for a numeric offset of the form +NN00 or -NN00, and the only hits are those four, the +0000 headers this page treats as UTC, and two MIME boundary strings of the form --00000000000… that the offset pattern matches by accident; the fourteen further sheets read for this pass carry no date line at all.^61533946

Field states. The second sheet of the January 2015 anchor opens with a redaction bar roughly two lines deep, measured at 600 dpi against the thin printed rule at the head of the first sheet, 212 pixels against 25, and the attribution line it covers is not resolved from that document.²¹ The To: line of the 10 January 2017 forward, the From: line of the 7 January 2014 schedule, and several personal names inside the 12 February 2018 and 2 December 2018 messages are likewise barred by the producing party, and none of them is supplied.^3864958 The subject line of the 3 January 2015 anchor is printed in the clear and reads “Re: today’s papers”.¹ Two sheets read here carry no covering at all: the 4–5 October 2011 anchor and the 19 August 2013 memo.⁷¹¹ No signature, printed or manuscript, appears on any sheet read for this page: these are e-mail exports, and the question does not arise.

Where the text layer lies on these particular sheets, for anyone re-deriving this page. Fourteen places were found where the extraction of a sheet cited here differs from the sheet, by literal substring tests against the rows followed by a render; on each one the page follows the image.

1. The extraction of the October 2011 anchor reads the addressee as Michael Woltflmichael©burnrate.corn; the sheet prints the name correctly.⁷
2. The extraction of the 5 October 2011 sheet reads a shortened link as tinyurl.corn, an rn read as m; the image prints tinyurl.com.¹⁹
3. The extraction of the February 2015 attachment name reads NYM--JE_Istf; the image prints NYM--JE_1.rtf.²⁵
4. The extraction of the second sheet of the January 2015 anchor reorders the forwarded message, lifting its header block into the middle of the body, where the image keeps header and body apart.²¹
5. The extraction of the first sheet of that anchor lifts “=omment?” above the sentence it ends.¹
6. The extraction of the July 2011 sheet reads weekend„ where the sheet prints weekend,,, two commas.¹⁵
7. The extraction of the April 2017 sheet reads speak =0 him where the sheet prints speak =o him, a digit for a letter inside a soft break.³²
8. The extraction of the February 2019 script thread reads “how i ye been mistreated” where the image prints “i ve”.⁶⁰
9. The extraction of the May 2019 book thread reads “3:30 is ood.” where the image prints “3:30 is good.”⁶¹
10. The extraction of the September 2011 acknowledgement reads “great lint” where the sheet prints “great fun”, which is why a phrase search for that sentence does not return the copy this page cites.⁶²
11. The extraction of the February 2015 photo-shoot sheet reads “Ma azine” and “we’re lannin” where the sheet prints “Magazine” and “we’re planning” in undamaged type.²⁷
12. The extraction of the April 2017 invitation reads the spelling as “michael wolfc”; at 600 dpi the sheet prints “michael wolfe”, the final letter carrying the horizontal bar of an e. The same extraction reads “Hello Teije” where the sheet prints “Hello Terje”. Brad Karp’s page cites the same sheet and now gives the same spelling; both pages follow the image.³⁵
13. The extraction of the House Oversight copy of the December 2015 advice drops the full stop after “hang himself” that the EFTA image prints. The House Oversight sheet has now been read at the image, on 2026-09-14, and it prints the full stop too: I think you should let him hang himself. So this is an ordinary extraction fault like the other thirteen, and the comparison is now image against image. An earlier version of this entry said that sheet “cannot be rendered” and that “nothing is claimed about the face of the House Oversight sheet”; the 403 is a fact about the /pdf/ route, the sheet is an ordinary page image, and the implied difference between the two copies dissolves at the image.⁸⁹
14. The extraction of the July 2018 sheet about meeting Starr reads the deepest quoted attribution line as “On The, Jul 10”; at 400 dpi the sheet prints “On Tue, Jul 10”, as its two other attribution lines do. The same extraction reads the address domain on that line as grnail.com, an rn read as m. An earlier version of this page gave the extraction’s “On The” as the sheet’s own face, in a footnote claiming image verification. It is the one fault of that kind found on this page, and the reason the sweep described in the note on the documents was run.⁵⁵

On all fourteen an automated comparison of the page against the text layer marks the page’s correct wording as a mismatch, because such a comparison measures agreement with the extraction rather than with the sheet.

One place the text layer does not lie, and it is worth stating because the opposite was reported here. The extraction of the August 2013 memo is complete: the pages row runs to 2,946 characters and carries the memo entire, A to F, the offer and all five closing paragraphs, and the row for the sibling copy at EFTA00968035 is 2,942 characters and likewise complete. Read through a print window that stops at about 1,600 characters, it appears to end inside option D at “…define a new normal by redefin”, which is character 1,614 of 2,946. Check a row’s length before reporting that it stops.¹¹²⁰

And many of these sheets carry quoted-printable soft breaks on their face: willing=ess, a=legations, has='t, reg=me, N=C, wh=re, =ea, Happy New Yea=, game-c=anging and in =ouch, which are what the documents print and are reproduced here unrepaired.

Identity numbers, account numbers, telephone and fax numbers, e-mail addresses, residential addresses and dates of birth printed on these sheets are not reproduced on this page.

Related

• People: Jeffrey Epstein; Ian Osborne; Brad Karp; Lesley Groff; Kathryn Ruemmler; Darren Indyke; Steve Bannon; Julie K. Brown; Thorbjørn Jagland; Paul Hoffman; Donald Trump Jr.; Mike Pompeo; Miroslav Lajčák; Harry Beller; Karyna Shuliak; Tom Pritzker; Terje Rød-Larsen
• Topics: Filthy Rich; Operation Leap Year

Sources

Footnotes

1. Michael Wolff to “jeffrey E.”, Sent: field “Saturday, January 3, 2015 8:35 PM”, subject “Re: today’s papers”. Four sheets. The advice quoted above is the outermost message on the first sheet; the sender’s address is barred in the header. Read at the image at 400 dpi. https://epstein-data.com/EFTA01791807 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

2. Michael Wolff to Jeffrey Epstein, Sent: field “Friday, May 3, 2013 4:52 PM”, subject “thanks…”. One sheet; the body is a single line continuing the subject, and the sender’s address is barred. Read at the image at 400 dpi. https://epstein-data.com/EFTA01758104 p.1 ↩ ↩² ↩³

3. Michael Wolff to Jeffrey Epstein, header Date: Tue, 08 Nov 2011 10:37:17 +0000, subject “Re:”, carrying beneath it Epstein’s message of “Nov 8, 2011, at 5:29 AM”, Wolff’s of “Tue, Nov 8, 2011 at 11:19 AM” signed “mw”, Epstein’s of “Nov 7, 2011, at 7:51 AM” reading “why dont you send ,me a financial proposal.”, Wolff’s of “Mon, Nov 7, 2011 at 11:31 AM” and Epstein’s of “Nov 7, 2011, at 5:27 AM”. Two sheets, the second carrying only the confidentiality footer and read at the image for this pass. The quoted clocks do not reconcile on any one zone and no interval is asserted between them; the order given in the body is the order of the nesting. The sender’s address is barred at each attribution line. Read at the image at 400 dpi. https://epstein-data.com/EFTA00719949 p.1 ↩ ↩² ↩³

4. Lesley Groff to Michael Wolff, header Date: Wed, 24 Jan 2018 14:34:22 +0000, subject “Re: Jeffrey Epstein”, carrying the whole morning’s exchange. Both parties’ addresses are barred, and the name of the person who was to collect the books is covered by a bar inside Groff’s own sentence and is not supplied here. Read at the image at 400 dpi. https://epstein-data.com/EFTA00467927 p.1 ↩ ↩²

5. Lesley Groff, circulated from Epstein’s office, Sent: field “Wed 1/22/2014 6:41:07 PM”, subject “Dinner on Sun 26th at RESTAURANT!”, carrying a fifteen-name list under the italic heading “Guest List at the Moment:”. Michael Wolff is the fourteenth name on it, and the line immediately below his carries two bars. Most of the recipient addresses in the To: line are barred. Read at the image at 400 dpi. https://epstein-data.com/EFTA02116547 p.1 ↩ ↩² ↩³

6. Daily schedule for Thursday 9 January 2014, circulated from Epstein’s office to Jeffrey Epstein and others, Sent: field “Tue 1/7/2014 4:17:20 PM”, subject “Josh Harris”. The sender’s line and most of the recipient addresses are barred, as is a contact value in parentheses after the Wolff entry. The day’s entries are set in italic type under a bold, italic, underlined date heading, with the times in bold; the sheet prints “Reminer”. Two sheets. The first carries the Wolff breakfast and six later appointments; the second carries three more, at 6:00pm, 7:00pm and 8:00pm, with values barred after each. The second sheet was opened only on the third pass, and the count on this page was a count of the first sheet until then. Read at the image at 400 dpi. https://epstein-data.com/EFTA01939225 pp.1-2 ↩ ↩² ↩³ ↩⁴

7. Ian Osborne to Jeffrey Epstein and Michael Wolff, copy to Jeffrey Epstein, Sent Wed 10/5/2011 4:19:44 AM, subject “Re: Re:”, carrying beneath it Wolff’s message quoted as “On 4 Oct 2011, at 05:57” and Epstein’s quoted as “On Oct 4, 2011, at 7:12 AM”. One sheet; nothing on it is covered by the producing party, and no passage continues onto a later sheet. The corpus extraction of the To: line reads “Michael Woltflmichael©burnrate.corn”; the image prints the name correctly. Read at the image at 400 dpi. https://epstein-data.com/EFTA01987731 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

8. Michael Wolff to “jeffrey E.”, Sent: field “Wednesday, December 16, 2015 4:27 PM”, subject “Re: Heads up”, quoting Epstein’s message of “Tue, Dec 15, 2015 at 11:52 PM” beneath it. This is the copy the quotation above is taken from, and it carries quoted-printable soft breaks throughout which are reproduced unrepaired. The quoted “Tue, Dec 15, 2015 at 8:00 PM” message’s body has dropped out of this copy, so only the boilerplate follows it. Read at the image at 400 dpi. https://epstein-data.com/EFTA02346409 p.1 ↩ ↩² ↩³ ↩⁴

9. The same message in the House Oversight production, Sent: field “12/16/2015 4:26:32 PM”. One sheet. Read at the image on 2026-09-14, and the sheet prints “I think you should let him hang himself.” with the full stop, in undamaged type, over the quoted exchange of “On Tue, Dec 15, 2015 at 11:52 PM” and “On Tue, Dec 15, 2015 at 8:00 PM”. The missing stop is this copy’s extraction, not this copy’s sheet. In that extraction the sentence reads “let him hang himself If he says”, with no full stop, and it carries none of the quoted-printable damage the EFTA copy shows; the EFTA sheet at ⁸ prints “let him hang himself.” with the stop, read at the image. So the two copies agree at the image and the comparison is image against image. An earlier version of this footnote said the sheet “cannot be rendered by any agent here and nothing is asserted about its face”. The /pdf/ endpoint does return HTTP 403 — re-measured 11 September 2026, against 200 for both EFTA copies — and that is a fact about the route, not about the sheet; the retraction of the rest of that sentence is the point of this correction. https://epstein-data.com/HOUSE_OVERSIGHT_032229 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

10. Michael Wolff to Jeffrey Epstein, header Date: Mon, 13 Jun 2011 14:25:47 +0000, subject “Re:”, quoting Epstein’s two-word message “wed thurs?” beneath it. One sheet; the sender’s address is barred and nothing else on the sheet is covered. The sheet prints “I’m leave” and “St. Petersberg”. Read at the image at 400 dpi. https://epstein-data.com/EFTA00913258 p.1 ↩ ↩² ↩³ ↩⁴

11. Jeffrey Epstein to Michael Wolff, header Date: Mon, 19 Aug 2013 23:50:38 +0000, subject “Re: Some thoughts”, carrying beneath his one-line reply the whole of Wolff’s message of “Mon, Aug 19, 2013 at 4:01 PM”. Two sheets, the second carrying only the confidentiality footer and read at the image for this pass. Wolff’s address is barred; nothing else on the first sheet is covered. The lettered heads A) to F) are on the face of the sheet and are not this page’s formatting. The pages row for this sheet is 2,946 characters and carries the memo entire, A to F, the offer and all five closing paragraphs. The surname “Gates” is printed bare and given no forename. Read at the image at 400 dpi. https://epstein-data.com/EFTA00968026 p.1 ↩ ↩² ↩³ ↩⁴

12. Michael Wolff to Jeffrey Epstein, header Date: Wed, 25 Sep 2013 10:44:28 +0000, subject “Re:”, quoting Epstein’s message of “Tue, Sep 24, 2013 at 4:42 PM” beneath it. One sheet; the sender’s address is barred. The sheet prints “didn;t” and “something to today”. Read at the image at 400 dpi. https://epstein-data.com/EFTA00689569 p.1 ↩ ↩² ↩³ ↩⁴

13. The six sheets whose Sent: field and Apple date-received integer were compared are ¹⁵ (1311184087 = 17:48:07 UTC, 20 July 2011, against a field of 5:48 PM), the sheet at https://epstein-data.com/EFTA02511614 p.1 (1422738859 = 21:14:19 UTC, 31 January 2015, against a field of 9:14 PM; “jeffrey E.” to Michael Wolff, subject “Re: hey…”, read at the image at 400 dpi), ³⁰ (1450241545 = 04:52:25 UTC, 16 December 2015, against a field of 4:52 AM), ⁴⁴ (1517595331 = 18:15:31 UTC, 2 February 2018, against a field of 6:16 PM), ⁴⁵ (1517595852 = 18:24:12 UTC, 2 February 2018, against a field of 6:24 PM) and ⁴⁷ (1528329737 = 00:02:17 UTC, 7 June 2018, against a field of 12:02 AM). Five of the six agree to the minute; ⁴⁴ does not. There the integer falls twenty-nine seconds before the minute the field prints, so on that one sheet the two agree to the hour and no further. Twenty-nine seconds does not bear on the four-hour question these comparisons exist to settle, but it is not agreement to the minute and an earlier version of this footnote said it was. All six sheets were read at the image for this page, at 400 dpi and ⁴⁷ again at 600, and the integers were converted with the Python standard library. A seventh sheet read at the image, ⁶³, also carries an integer, but it is ninety seconds off its field and is not relied on here. ↩ ↩²

14. Searches run with the FTS5 index over /mnt/d/dump/AI-investigator/Epstein-research-data/full_text_corpus.db on 10 September 2026, and the absence tests and per-production shares re-derived on 11 September 2026, read-only and immutable, with no LIKE scan, no LIMIT on any counting query, and row identifiers resolved only after selection. Page counts are rows of pages_fts; document counts are rows of doc_search_fts. doc_search.search_text is capped at 300 characters, so every document figure on this page counts documents whose first few hundred characters carry the phrase, not documents containing it anywhere; on the independent basis of distinct documents over the matching pages, "Michael Wolff" is 5,478 documents rather than 5,388, and the ten-form union differential is 26 of 5,478, 0.47%, against 25 of 5,388, 0.46% — the same answer on either basis, which is the strongest thing that can be said for it. The damaged-spelling forms were generated by python3 scripts/ocr_variants.py "Michael Wolff" --wide --count, which produced 37 forms of which 11 returned non-zero; each non-zero form was sampled at the rows before it was accepted or rejected. The per-production shares quoted for the one-f form are of pages, computed as the damaged-only page count in each production over that production’s canonical page count plus its damaged-only count, and all four are on that one basis. House Oversight is scoped by identifier prefix and not by the documents.dataset column. Re-derived 11 September 2026: of the 331 canonical HOUSE_OVERSIGHT_ pages, every one has a row in the documents table, and 87 of those rows carry dataset NULL while 244 carry 99. A scoping through dataset = 99 therefore keeps 244 of the 331 and sweeps in two DOJ-OGR pages, giving 246 and a share of 6/252 = 2.38%, where the identifier prefix gives 331 and 6/337 = 1.78%. The defect is a null column, not a missing row, so it is not the same defect as the row noted below which the index holds under a name that is not a production number and which has no documents entry at all; and a diagnostic that tests for a missing row returns zero on all 331 of these. The corpus-wide figure, measured the same day, is 8,717 of 30,507 HOUSE_OVERSIGHT_ rows with dataset NULL, all of them carrying a valid total_pages. The homograph families were each confirmed by a separate document query and checked to lie inside the 128-document residual. The December 2015 thread was tested by its own quoted timestamp strings rather than by phrases from its text, because a search built from wording this page already quotes cannot disconfirm this page: "Dec 15, 2015 at 8:00 PM" returns 4 pages, "Dec 15, 2015 at 11:52 PM" 3, "12/16/2015 1:00:36" 1 and "12/16/2015 4:26:32" 2, and their union is five pages. One of the five is a row the index holds under the name “3-emails_redacted_p3”, which is not a production number and has no entry in the document table, and which is not cited anywhere on this page. The sixteen report-back forms, the twenty-six acknowledgement forms and the twenty further acknowledgement forms run on 11 September 2026 are listed in full in the body above, with what each returned. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰

15. Michael Wolff to Jeffrey Epstein, 20 July 2011, subject “Re:”, Sent: field “Wednesday, July 20, 2011 5:48 PM”, quoting Epstein’s message beneath it. The sender’s e-mail address is barred on this sheet; the sheet also prints an Apple property list in which date-received reads 1311184087, which is 17:48:07 UTC on 20 July 2011 and fixes the Sent: field as UTC. Read at the image at 400 dpi. https://epstein-data.com/EFTA02691062 p.1 ↩ ↩² ↩³

16. Ian Osborne to Jeffrey Epstein, copy to Michael Wolff, header Date: Wed, 15 Jun 2011 16:01:31 +0000, subject “Meeting”, with an attachment the sheet names “Jeffrey_Epstein_110613.pdf”. Cited from the corpus text layer and not rendered; the message is Osborne’s rather than Wolff’s and is used here only for its first sentence and its date. UNVERIFIED. https://epstein-data.com/EFTA00719271 p.1 ↩ ↩²

17. Michael Wolff to Jeffrey Epstein, copy to Ian Osborne, header Date: Tue, 04 Oct 2011 12:57:18 +0000, subject “Re:”, quoting Epstein’s message of “Oct 4, 2011, at 7:12 AM” beneath it. Wolff’s and Osborne’s e-mail addresses are barred on this sheet. Read at the image at 400 dpi. https://epstein-data.com/EFTA00695056 p.1 ↩

18. The same Osborne message produced separately, with an explicit Date: Wed, 05 Oct 2011 04:19:44 +0000 header in place of the unlabelled Sent: field. Read at the image at 400 dpi. https://epstein-data.com/EFTA00720109 p.1 ↩ ↩²

19. Michael Wolff to Ian Osborne, copy to Jeffrey Epstein, header Date: Wed, 05 Oct 2011 11:22:52 +0000, subject “Re: Re:”, carrying the whole earlier chain beneath it. The corpus extraction reads the shortened link as “http://tinyurl.corn/43enzct”; the image prints “http://tinyurl.com/43enzct”. Read at the image at 400 dpi. https://epstein-data.com/EFTA00921553 p.1 ↩ ↩²

20. Michael Wolff to Jeffrey Epstein, header Date: Tue, 20 Aug 2013 00:51:07 +0000, subject “Re: Some thoughts”, the whole body “my pleasure”, carrying beneath it Epstein’s message quoted as “On Mon, Aug 19, 2013 at 7:50 PM” and, beneath that, the whole of Wolff’s memo of “On Mon, Aug 19, 2013 at 4:01 PM” from “An interesting predicament…” through options A to F, the offer and the closing paragraphs. Wolff’s address is barred in the header. The interval of one hour, no minutes and twenty-nine seconds is measured between this header and the Date: header of ¹¹, both explicit +0000: 23:50:38 to 00:51:07. Two sheets: the memo’s last three lines — “Anyway, in a nutshell.”, “I leave for Berlin on Sunday and then London. I’m back on September 7.” and the sign-off “m” — stand on the second, above the boilerplate. Read at the image at 400 dpi. https://epstein-data.com/EFTA00968035 pp.1-2 ↩ ↩²

21. Same document, its own p. 2 of 4: a redaction bar at the head of the sheet covering the attribution line, then the message beginning “Can you give me time middle of the month”, then Epstein’s forward of “On Sat, Jan 3, 2015 at 6:19 AM” with the word “whoops” and Ian Osborne’s forwarded message of “Sat, Jan 3, 2015 at 5:38 AM”. The bar was measured on a 600 dpi render at 212 pixels deep against the 25-pixel printed rule at the head of p. 1, which is how it was distinguished from a rule; only descender tips survive below its lower edge and nothing under it is resolved from this document. Read at the image at 400 dpi and again at 600 dpi. https://epstein-data.com/EFTA01791807 p.2 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

22. Michael Wolff to “jeffrey E.”, Sent: field “Saturday, January 3, 2015 2:14 PM”, subject “Re: today’s papers”, the same message that EFTA01791592 quotes as “On Sat, Jan 3, 2015 at 10:14 AM”. This sheet prints “schedule Gate=”, with the = a quoted-printable soft break; EFTA01791592 prints “schedule Gates”. The four-hour difference between the Sent: field and the quoted line is measured from these two sheets. Read at the image at 400 dpi and again at 600 dpi to place the soft breaks. https://epstein-data.com/EFTA01790536 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵

23. Michael Wolff to “jeffrey E.”, Sent: field “Saturday, January 3, 2015 3:39 PM”, subject “Re: today’s papers”, three sheets, whose first sheet carries the whole of the morning’s exchange with each quoted message’s own clock. The nesting on this sheet is branched rather than strictly chronological. The order given in the bulleted list above is the order of the clocks and is not the order of the sheet; the sheet’s own order is set out in the paragraph beneath that list, and was read at the image. No clock of 11:38 appears anywhere on this sheet: the text the list gives at that clock stands here as the body of the sheet’s own 3:39 PM message. Several e-mail addresses in the quoted attribution lines are barred. Read at the image at 400 dpi and again at 600 dpi to place the soft breaks. https://epstein-data.com/EFTA01791592 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

24. Same document, its own pp. 3 and 4 of 4: repeated copies of the confidentiality footer and nothing else. No further message, qualification or reply stands on either sheet. Read at the image at 300 dpi, which is enough to settle whether any message stands on a sheet. https://epstein-data.com/EFTA01791807 pp.3-4 ↩ ↩²

25. Michael Wolff to “jeffrey E.”, header Date: Thu, 05 Feb 2015 12:30:58 +0000, subject “Re:”, attachment named on the sheet as “NYM—JE_1.rtf”, quoting Epstein’s message of “Thu, Feb 5, 2015 at 6:40 AM” beneath it. The corpus extraction reads the attachment name as “NYM—JE_Istf”. Read at the image at 400 dpi. https://epstein-data.com/EFTA01191300 p.1 ↩ ↩² ↩³

26. Michael Wolff to Jeffrey Epstein, Sent: field “Tuesday, February 10, 2015 7:22 AM”, subject “Fwd: Jeffrey Epstein / portrait”, forwarding his own exchange with a magazine’s photo desk. Cited from the corpus text layer and not rendered; verifying it would cost one page render, and the sheet carries quoted-printable damage which is reproduced here unrepaired. UNVERIFIED. https://epstein-data.com/EFTA02400213 p.1 ↩ ↩²

27. Michael Wolff to a recipient whose address the production bars, Sent: field “Fri 2/20/2015 6:30:15 PM”, subject “Re: Hey Lesley”; the body is one word, “Fab!”, above a quoted message whose attribution line is barred as to the name and address only, and, beneath that, Wolff’s own message quoted as “On Feb 20, 2015, at 1:18 PM”. A bar stands inside the quoted 1:18 PM message after the contact name. The corpus extraction reads “Ma azine” and “we’re lannin”; the sheet prints “Magazine” and “we’re planning” in undamaged type. Read at the image at 400 dpi. https://epstein-data.com/EFTA02085392 p.1 ↩ ↩²

28. Michael Wolff to “Jeffrey E.”, header Date: Fri, 13 Feb 2015 07:51:00 +0000, subject “Re:”, answering Epstein’s message of “Fri, Feb 13, 2015 at 12:39 AM” carrying a news link. Cited from the corpus text layer and not rendered. UNVERIFIED. https://epstein-data.com/EFTA00709562 p.1 ↩ ↩²

29. Michael Wolff to Jeffrey Epstein, Sent: field “12/16/2015 1:00:36 AM”, subject “Heads up”, “Importance: High”. One sheet. Read at the image on 2026-09-14, where the header block prints From: Michael Wolff with the address barred, To: Jeffrey Epstein [[email protected]], and the whole body is the single line I hear CNN planning to ask Trump tonight about his relationship with you--either on air or in scrum afterwards. — exact. An earlier version of this footnote said the sheet “cannot be rendered”. The /pdf/ route does return HTTP 403 for HOUSE_OVERSIGHT_* identifiers; the sheet is an ordinary page image and reads. The same message is quoted in full on EFTA02477268, which was read at the image. https://epstein-data.com/HOUSE_OVERSIGHT_032228 p.1 ↩ ↩² ↩³ ↩⁴

30. “jeffrey E.” to Michael Wolff, Sent: field “Wednesday, December 16, 2015 4:52 AM”, subject “Re: Heads up”, quoting Wolff’s message of “Tue, Dec 15, 2015 at 8:00 PM” in full. The sheet prints an Apple date-received integer of 1450241545, which is 04:52:25 UTC on 16 December 2015 and fixes the Sent: field as UTC; on that footing the quoted 8:00 PM of 15 December is the same instant as the 1:00:36 AM UTC Sent: field of HOUSE_OVERSIGHT_032228. Wolff’s address is barred in the quoted attribution line. Read at the image at 400 dpi. https://epstein-data.com/EFTA02477268 p.1 ↩ ↩² ↩³

31. Michael Wolff to “jeffrey E.”, Sent: field “Thursday, March 9, 2017 3:17 PM”, subject “Re:”, quoting Epstein’s message of “Saturday, March 4, 2017” beneath it. The sender’s address is barred. The sheet prints “N=C.” and “there’=”, “ch=t”, “o=her”, “wh=re” and “wi=l”, each a quoted-printable soft break carried onto the face of the document. Read at the image at 400 dpi. https://epstein-data.com/EFTA02353297 p.1 ↩ ↩²

32. Michael Wolff to Jeffrey Epstein, Sent: field “Sunday, April 23, 2017 8:15 PM”, subject “PR”. One sheet, carrying one line of body text and nothing else; the sender’s address is barred. The surname prints “Hiltzig”: read at the image at 400 dpi and the final letter re-read at 600 dpi, where it carries a descender. https://epstein-data.com/EFTA02366035 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵

33. Michael Wolff to Jeffrey Epstein, header Date: Tue, 04 Dec 2018 13:34:27 +0000, subject “Re:”, carrying the whole morning’s exchange beneath it. The sheet prints “post-morten” and “Hiltzig” as quoted. The sender’s address is barred at both attribution lines. Read at the image at 400 dpi. https://epstein-data.com/EFTA01014465 p.1 ↩ ↩² ↩³

34. Lesley Groff to Jeffrey Epstein, copy to Rich Kahn, header Date: Thu, 04 May 2017 13:54:57 +0000, subject “Paul Barrett”, setting out the schedules for Thursday 4 and Friday 5 May 2017 under bold, italic, underlined date headings, with the entries in italic, the time in bold at the head of each and, where an entry opens with a capitalised meal word, that word bold as well. Two entries are set wholly bold: “6:30pm Doors Open for Terje’s Opening of: Oslo” on the first sheet and “7:00pm Performance of Oslo” on the second. Every address value in the From:, To: and Cc: lines is barred, and a value in parentheses is barred after several entries including the Wolff lunch. The sheet prints “Jefffrey” in the To: line, spells the surname “Hiltzik” with a final k, and prints “Opening of: Oslo” where the extraction reads “Opening of.• Oslo”. Two sheets. The second carries only six italic lines for the 7:00pm performance, the first of them wholly bold and the last “Jojo to drive you!!!”, and nothing on it is covered by the producing party. The first sheet was read for the second pass and the second sheet only in the third, which is why a schedule described here as read at the image was for a time read only in part. Read at the image at 400 dpi. https://epstein-data.com/EFTA00450519 pp.1-2 ↩ ↩² ↩³

35. A message to Terje Rød-Larsen, Sent: field “Fri 4/21/2017 11:42:00 AM”, subject “Re:”, reading “Hello Terje. Email addresses below:” above three lines — “Brad Karp.”, “Sami daughter is his assistant” and “Michael Wolff” — each followed by a barred address value. The From: line of that message is barred in full and this page does not identify its sender. Beneath it the sheet quotes Terje Rød-Larsen’s message of “On Apr 21, 2017, at 5:17 AM” and, beneath that, Epstein’s of “On Apr 20, 2017, at 22:42” reading “can we invite brad karp and michael wolfe to oslo”. Rød-Larsen’s surname is given in the To: line and in the quoted attribution line, spelled “Rod-Larsen”; his address value is barred. The final letter of “wolfe” was read at 600 dpi and carries the horizontal bar of an e. The word “oslo” is printed lower case and is not expanded on this page. Read at the image at 400 dpi and again at 600 dpi. https://epstein-data.com/EFTA02210485 p.1 ↩ ↩² ↩³

36. Michael Wolff to “jeffrey E.”, header Date: Sat, 06 May 2017 13:40:25 +0000, subject “Re:”, the whole body quoted above, carrying beneath it Epstein’s message quoted as “On Fri, May 5, 2017 at 11:38 PM”. One sheet; Wolff’s address is barred in the From: line and nothing else on the sheet is covered. The quoted clock runs four hours behind the explicit +0000 header of the message it quotes and no zone is asserted for it. A second copy at https://epstein-data.com/EFTA02389232 p.1, read in the corpus text layer only, carries the same message with a Sent: field of “Saturday, May 6, 2017 1:40 PM” and quoted-printable damage reading “rea=ly liked” and “11=38 PM”. UNVERIFIED as to that copy. Read at the image at 400 dpi. https://epstein-data.com/EFTA00696039 p.1 ↩ ↩² ↩³ ↩⁴

37. “jeffrey E.” to Michael Wolff, header Date: Sat, 06 May 2017 03:38:04 +0000, the whole body “hope you guys had fun”. One sheet; the Subject: label prints with no value after it and no bar, so the field is blank rather than withheld, and Wolff’s address is barred. This is the message quoted beneath the acknowledgement at ³⁶, surviving separately with its own explicit +0000 header. Read at the image at 400 dpi. https://epstein-data.com/EFTA01047230 p.1 ↩ ↩²

38. “jeffrey E.” forwarding Michael Wolff’s message of “Tue, Jan 10, 2017 at 8:31 AM”, Sent: field “Tue 1/10/2017 1:32:45 PM”, subject “Fwd: Ailes”, carrying the whole morning’s exchange beneath it. The To: line of the forward is barred in full and the recipient is not identified here; Wolff’s address is barred at each attribution line. Read at the image at 400 dpi. https://epstein-data.com/EFTA02199945 p.1 ↩ ↩² ↩³

39. Michael Wolff to Jeffrey Epstein, Sent: field “Monday, October 9, 2017 1:18 PM”, subject “Fwd: How’s it looking…?”, forwarding his exchange with Steve Bannon of 8 October 2017 in full. Two sheets: the second carries the earliest message the forward preserves, Wolff’s own, under the header Date: Sun, 8 Oct 2017 09:42:39 -0400, subject “How’s it looking…?”, body “lemme know”. That header is the only numeric zone offset on any sheet read for this page. Every display address on both sheets is barred. The second sheet was opened only on the third pass. Read at the image at 400 dpi. https://epstein-data.com/EFTA02589079 pp.1-2 ↩ ↩²

40. “jeffrey E.” to Michael Wolff, Sent: field “Wednesday, November 29, 2017 3:53 PM”, subject “Re: sulzberger?”, quoting Wolff’s message beneath it. Cited from the corpus text layer and not rendered; the sheet carries quoted-printable damage which is reproduced unrepaired. UNVERIFIED. https://epstein-data.com/EFTA02561344 p.1 ↩ ↩²

41. Michael Wolff to “jeffrey E.”, header Date: Fri, 05 Jan 2018 11:04:21 +0000, subject “Re:”, quoting Epstein’s message of “Thu, Jan 4, 2018 at 8:41 PM” beneath it. The sender’s address is barred. Read at the image at 400 dpi. https://epstein-data.com/EFTA00916741 p.1 ↩

42. “jeffrey E.” to Kathy Ruemmler, Sent: field “Thursday, January 4, 2018 3:29 PM”, quoting her message of “Thu, Jan 4, 2018 at 10:27 AM” beneath a trade-press link. Cited from the corpus text layer and not rendered; the sheet carries quoted-printable damage which is reproduced unrepaired. UNVERIFIED. https://epstein-data.com/EFTA02544962 p.1 ↩ ↩²

43. Michael Wolff to Jeffrey Epstein, Sent: field “Friday, February 2, 2018 4:11 PM”, subject “A couple of things…”, three short paragraphs of which the first is quoted here in full. Wolff’s address is barred. The sheet prints “in =ouch”, “discu=s” and “Fired=and Furious”, each a quoted-printable soft break carried onto the face of the document. Read at the image at 400 dpi. https://epstein-data.com/EFTA02533377 p.1 ↩ ↩² ↩³

44. “jeffrey E.” to “Brad S Karp; Michael Wolff”, Sent: field “Friday, February 2, 2018 6:16 PM”, the whole body “brad- michael , micahel brad”. One sheet. Both recipients are named in the clear, no address value is printed for either, and nothing on the sheet is covered by the producing party. There is no Subject: label on this sheet at all, which is a different field state from the blank labels elsewhere on this page. The sheet carries a metadata line in which date-received reads 1517595331, which is 18:15:31 UTC on 2 February 2018. The Sent: field reads 6:16 PM, and the two do not agree to the minute: the integer falls twenty-nine seconds before the minute the field prints, and this is the only one of the six sheets compared at ¹³ where they do not match minute for minute. The boilerplate beneath the body prints “=he information”, “=ttorney-client”, “in=ended”, “p=rt”, “rece=ved”, “ret=rn”, “an= all” and “r=served”, each a quoted-printable soft break. Read at the image at 400 dpi. https://epstein-data.com/EFTA02532742 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵

45. “Karp, Brad S” to “jeffrey E.” and Michael Wolff, Sent: field “Friday, February 2, 2018 6:24 PM”, subject “RE:”, the body quoted here in full above the line “Sent with BlackBerry Work”. Beneath it the sheet quotes Epstein’s message with a Date: line of “Friday, Feb 02, 2018, 1:15 PM”, a blank subject line and the whole body “brad- michael , micahel brad”. Karp’s and Wolff’s address values are barred in both the header and the quoted lines, and no value is taken from this sheet. The sheet carries a metadata line in which date-received reads 1517595852, which is 18:24:12 UTC on 2 February 2018 and agrees to the minute with the Sent: field of 6:24 PM — a further instance of the zone finding set out below. Read at the image at 400 dpi. https://epstein-data.com/EFTA02533200 p.1 ↩ ↩² ↩³ ↩⁴

46. Lesley Groff to Jeffrey Epstein, header Date: Tue, 30 Jan 2018 18:17:48 +0000, subject “Fwd: Leon”, carrying the day sheet for “Wed. Jan. 31, 2018 NY” above a forwarded message. The day’s entries are set in italic type under a bold, italic, underlined date heading, with the time in bold at the head of each entry and, where the entry opens with a capitalised meal word, that word bold as well: the dinner entry sets “7:00pm DINNER” bold and the remainder of the line in regular italic. Both display addresses in the header are barred, and a contact value in parentheses is barred after each of the three names in the dinner entry. Only the dinner entry is quoted on this page. The sheet prints “Jefffrey” in the To: line. Inside the Begin forwarded message: block the sheet carries the header “Date: January 30, 2018 at 1:07:31 PM EST”, which is one of the three explicit local-zone labels on any sheet read for this page. The same day sheet is cited on Miroslav Lajčák’s page. Two sheets: the second carries an Apollo Global Management confidentiality footer and nothing else, and was read at the image for this pass. Read at the image at 400 dpi. https://epstein-data.com/EFTA00468653 p.1 ↩ ↩² ↩³ ↩⁴

47. “jeffrey E.” to Michael Wolff, Sent: field “Thursday, June 7, 2018 12:02 AM”, subject “Re: Friday”, the whole body “whatever you want . before after ok with me.” plus, after a quoted-printable artefact printed as =C2 and one unprintable glyph, “anytime fine.”; above Wolff’s message quoted as “On Wed, Jun 6, 2018 at 7:41 PM”. Wolff’s address is barred in the quoted attribution line. The sheet prints “com=ng”, “Br=d”, “I.e=” and “will=be”, each a quoted-printable soft break. A metadata line at the foot of the sheet gives date-received as 1528329737, which is 00:02:17 UTC on 7 June 2018 and agrees to the minute with the Sent: field of 12:02 AM; it is the sixth of the six sheets compared at ¹³. The first character of “I.e=” is a bare vertical stroke, and the sheet’s typeface does not distinguish a capital I from a lower-case l: it is given here as the capital, which is how the sense reads, and the extraction reads it as the lower case. Read again at 600 dpi and measured there: the stroke is fifty-six pixels tall and six wide; the capital I of “I don’t”, on the same line, is fifty-six by seven; and the lower-case l of “talk”, also on that line, is fifty-six by six. The stroke matches a known lower-case letter exactly and the capital to within one pixel of width, so the image does not settle it and a higher resolution will not: the sheet is a raster of about one hundred dots per inch, so a 600 dpi render is a sixfold upscale of the same pixels and adds no information. The initials in the quoted message are not expanded on this page. Read at the image at 400 dpi. https://epstein-data.com/EFTA02646476 p.1 ↩ ↩² ↩³ ↩⁴

48. Michael Wolff to Jeffrey Epstein, Sent: field “Sat 2/8/2014 4:45:38 PM”, subject “Re:”, carrying beneath it Epstein’s message quoted as “On Sat, Feb 8, 2014 at 11:26 AM” and, beneath that, the whole of an exchange of 27 January 2014 whose last quoted message continues onto the second sheet. Two sheets. Wolff’s address is barred at every attribution line. The quoted clocks of the two men in the 27 January exchange do not reconcile on one zone and no interval is asserted between them. Read at the image at 400 dpi. https://epstein-data.com/EFTA01933332 pp.1-2 ↩ ↩²

49. “jeffrey E.” to Wolff, header Date: Mon, 12 Feb 2018 13:02:07 +0000, subject “Re: WA and SY”, carrying Wolff’s message of “Mon, Feb 12, 2018 at 7:59 AM”, Epstein’s of “6:50 AM” and Wolff’s of “Sun, Feb 11, 2018 at 5:42 PM”. The forename in the To: line is barred, as are three personal names inside the 11 February message; none is supplied here, and the 11 February message is not quoted. The initials in the subject line are not expanded. Read at the image at 400 dpi. https://epstein-data.com/EFTA00888548 p.1 ↩ ↩²

50. Michael Wolff to Jeffrey Epstein, header Date: Wed, 21 Feb 2018 13:56:54 +0000, subject “Fwd:”, forwarding a message from Juleanna Glover of “Wed, Feb 21, 2018 at 2:54 PM”. Wolff adds no text of his own above the forward; a short bar stands below the forwarded body. Read at the image at 400 dpi. https://epstein-data.com/EFTA00877624 p.1 ↩

51. “jeffrey E.” to “Ken Starr” and “Michael Wolff”, header Date: Sat, 19 May 2018 17:40:53 +0000, the whole body “Ken - Michael , Michael/ Ken”. One sheet; both recipients are named on the sheet and both address values are barred. The Subject: label prints with no value after it and no bar, so the field is blank rather than withheld. The same message survives at https://epstein-data.com/EFTA02660502 p.1, whose Sent: field reads “Saturday, May 19, 2018 5:41 PM”, read in the corpus text layer only. UNVERIFIED as to that copy. Read at the image at 400 dpi. https://epstein-data.com/EFTA01055521 p.1 ↩ ↩² ↩³

52. “Starr, Ken” to Michael Wolff, copy to Jeffrey Epstein, header Date: Sat, 19 May 2018 20:25:18 +0000, subject “Re:”, carrying beneath it Wolff’s message quoted as “On May 19, 2018, at 1:14 PM” and, beneath that, Epstein’s introducing message quoted as “On Sat, May 19, 2018 at 1:40 PM”. Starr’s surname is given in the From: line as the sheet sets it; his and Wolff’s address values are barred, and an address value inside the body is barred by the producing party and no value is taken from it. The two quoted clocks do not reconcile with the explicit +0000 headers on one zone and no zone is asserted for either; the interval given in the body above is measured between the +0000 header of ⁵¹ and the +0000 header of this sheet. Wolff’s message survives separately at https://epstein-data.com/EFTA02661102 p.1 and https://epstein-data.com/HOUSE_OVERSIGHT_033564 p.1, and Starr’s at https://epstein-data.com/EFTA02660675 p.1 and https://epstein-data.com/HOUSE_OVERSIGHT_032774 p.1, all read in the corpus text layer only. UNVERIFIED as to those four copies. Read at the image at 400 dpi. https://epstein-data.com/EFTA01055612 p.1 ↩ ↩² ↩³ ↩⁴

53. “Starr, Ken” to Jeffrey Epstein, Sent: field “Wednesday, June 13, 2018 3:43 PM”, subject “Fwd: Possible quote…”, forwarding Michael Wolff’s message to him of that morning under the header “Date: June 13, 2018 at 10:18:43 AM EDT”. One sheet. Starr’s and Wolff’s address values are barred. The proposed quotation is set on the sheet as an indented paragraph in the body of Wolff’s message and the wording above is taken from the image. Read at the image at 400 dpi. https://epstein-data.com/EFTA02629642 p.1 ↩ ↩² ↩³ ↩⁴

54. “Jeffrey E.” to Michael Wolff, header Date: Mon, 09 Jul 2018 22:32:12 +0000, subject “Re:”, the whole body “of course”, carrying beneath it Wolff’s message quoted as “On Mon, Jul 9, 2018 at 6:18 PM” and four earlier messages in the same exchange, at 6:17 PM, 6:13 PM, 6:07 PM and 6:05 PM. Wolff’s address is barred at each attribution line. Two sheets: the second carries the tail of the Epstein-house boilerplate and no message, and was read at the image for this pass. Read at the image at 400 dpi. https://epstein-data.com/EFTA01003826 p.1 ↩ ↩²

55. “Jeffrey E.” to Michael Wolff, header Date: Tue, 10 Jul 2018 21:50:44 +0000, subject “Re:”, the whole body “Great have fun”, carrying beneath it Wolff’s message quoted as “On Tue, Jul 10, 2018 at 5:49 PM” and two earlier messages. Wolff’s address is barred at each attribution line. All three On … attribution lines on the sheet print “On Tue”, the deepest of them included; it is the extraction that reads “On The, Jul 10” there, and an earlier version of this footnote gave the extraction’s form as the sheet’s. See item 14 of the text-layer list above. Read at the image at 400 dpi. https://epstein-data.com/EFTA01005209 p.1 ↩ ↩² ↩³

56. “jeffrey E.” to Michael Wolff, header Date: Sun, 26 Nov 2017 11:04:58 +0000, the whole body quoted above. One sheet; the Subject: label prints with no value and no bar, and Wolff’s address is barred. The same message survives at https://epstein-data.com/EFTA00946343 p.1 and https://epstein-data.com/EFTA02562127 p.1, read in the corpus text layer only. UNVERIFIED as to those two copies. Read at the image at 400 dpi. https://epstein-data.com/EFTA00946297 p.1 ↩ ↩²

57. Michael Wolff to Jeffrey Epstein, Sent: field “Friday, August 24, 2018 11:42 AM”, subject “Re:”, quoting Epstein’s message of “Fri, Aug 24, 2018 at 6:54 AM” beneath it. The sheet prints “a =ea this afternoon”, the = a quoted-printable soft break. Read at the image at 400 dpi. https://epstein-data.com/EFTA02608521 p.1 ↩

58. “J” forwarding Kathy Ruemmler’s message of “Sun, Dec 2, 2018 at 9:07 PM”, header Date: Mon, 03 Dec 2018 02:21:44 +0000, subject “Fwd: thoughts”, carrying Ruemmler’s message to Michael Wolff copied to Epstein and Darren Indyke, Wolff’s message of “Dec 2, 2018, at 9:02 PM”, and Epstein’s of “8:51 PM” beneath them. The forward’s own To: line is barred, as is one name inside Ruemmler’s sentence. Epstein’s message, which is not quoted in full here, is a list of points he thought reporters ought to make about his case. Read at the image at 400 dpi. https://epstein-data.com/EFTA01014613 p.1 ↩ ↩² ↩³

59. “J” to Michael Wolff, header Date: Fri, 01 Feb 2019 10:18:49 +0000, subject “Fwd:”, reading “have fun” above a message Epstein had sent to his own address at “Fri, Feb 1, 2019 at 5:18 AM”. Wolff’s address is barred, and one personal name inside the forwarded body is barred; neither is supplied here. The forwarded body is not quoted on this page beyond the phrase that names the investigation, which is set out in full on the topic page. Two sheets: the forwarded body continues onto the second, which was read in the corpus text layer only and carries no message of Wolff’s. UNVERIFIED as to the second sheet. Read at the image at 400 dpi. https://epstein-data.com/EFTA01028632 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵

60. “J” to Michael Wolff, header Date: Wed, 27 Feb 2019 14:18:31 +0000, subject “Re: script draft”, carrying the whole morning’s exchange beneath it. Wolff’s address is barred at each attribution line. Read at the image at 400 dpi. https://epstein-data.com/EFTA01032286 p.1 ↩ ↩²

61. Bella Klein to Lesley Groff, header Date: Tue, 28 May 2019 18:41:17 +0000, subject “Re: Jeffrey Epstein”, carrying Groff’s instruction of “May 28, 2019, at 2:39 PM” and, beneath it, the forwarded exchange with Michael Wolff running from 27 to 28 May 2019. Wolff’s message of 28 May carries the header “Date: May 28, 2019 at 2:13:14 PM EDT”, one of three explicit local-zone labels on any sheet read for this page. Every display address on the sheet is barred, and two short bars stand beneath two of the message bodies. The sheet prints “forst 40 pages”. Two sheets: the second carries the earliest message the forward preserves, Groff’s first request, and her signature block, which gives her own role. The second sheet was opened only on the third pass. Read at the image at 400 dpi. https://epstein-data.com/EFTA00496165 pp.1-2 ↩ ↩² ↩³ ↩⁴

62. “jeffrey E.” — the sheet gives the From: line as Jeffrey Epstein — to Michael Wolff, Sent: field “Fri 9/2/2011 7:36:59 PM”, subject “Re: thanks”, the whole body “i am thrilled that you looked so happy”, above Wolff’s message quoted as “On Fri, Sep 2, 2011 at 9:41 PM” reading “Well that was great fun. I loved that guy. And Victoria loved you. Soon. m”. Wolff’s address is barred in both the To: line and the quoted attribution line; nothing else on the sheet is covered. The quoted clock runs ahead of the Sent: field and no zone is asserted for it. The corpus extraction of this sheet reads “great lint” where the sheet prints “great fun”. Five further copies of Wolff’s message survive, at https://epstein-data.com/EFTA00918985 p.1, https://epstein-data.com/EFTA01854608 p.1, https://epstein-data.com/EFTA01854877 p.1, https://epstein-data.com/EFTA02543424 p.1 and https://epstein-data.com/EFTA02693390 p.1, read in the corpus text layer only; the quotation here is taken from the image of the sheet cited. The phrase search "was great fun" reaches only four of the six copies, because the extraction of this sheet and of EFTA01854877 both read “great lint”; the fragment "that was great", which stops short of the damaged word, returns all six. UNVERIFIED as to the five further copies. Read at the image at 400 dpi. https://epstein-data.com/EFTA01854645 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵

63. Michael Wolff to Jeffrey Epstein, Sent: field “Thursday, February 16, 2012 7:31 PM”, subject “Brooklyn”, the whole body “That was fun”, followed on the sheet by an Apple property list in which date-received reads 1329420570, which is 19:29:30 UTC on 16 February 2012 and is within two minutes of the field rather than equal to it, so it is not relied on here as a fourth zone check. Wolff’s address is barred. Epstein’s reply, “Agreed”, survives at https://epstein-data.com/EFTA02696647 p.1, read in the corpus text layer only. Read at the image at 400 dpi. https://epstein-data.com/EFTA02550537 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

64. Michael Wolff to Jeffrey Epstein, header Date: Mon, 27 Jan 2014 14:15:56 +0000, the whole body “Super! Really nice evening. Thanks. When are you back?” One sheet; the Subject: label prints with no value after it and no bar, so the field is blank rather than withheld; the sender’s address is barred. The same message survives in twenty-one further copies, among them https://epstein-data.com/EFTA01933332 pp.1-2, where it stands at the foot of a chain quoted in reverse order: the 9:15 AM message is the earliest in that chain and the exchange arranging a four o’clock visit follows it. EFTA01933332 was read at the image at 400 dpi; the other copies were read in the corpus text layer only. Read at the image at 400 dpi. https://epstein-data.com/EFTA00684779 p.1 ↩ ↩² ↩³

65. Michael Wolff to Jeffrey Epstein, Sent: field “Tuesday, March 6, 2018 1:52 PM”, subject “That was fun”, the whole body quoted here. Wolff’s address is barred; the sheet prints “=rom”, a quoted-printable soft break. Epstein’s one-word reply and Wolff’s answer to it survive at https://epstein-data.com/EFTA02512140 p.1 and https://epstein-data.com/EFTA02512309 p.1, read in the corpus text layer only and not quoted here. UNVERIFIED as to those two copies. Read at the image at 400 dpi. https://epstein-data.com/EFTA02511806 p.1 ↩ ↩²

66. An alphabetised roster of names with one-line biographies, held in the release under several production numbers. Its entry between “Katharine Weymouth” and “Paul Wolpe” reads “Michael Wolf – CEO, Activate. fmr President and COO, MTV Networks. Fmr Prtnr, McKinsey. Former Senior Partner, Booz Allen & Hamilton.” The surname is printed with one f in clean typeset text with no damage anywhere on the sheet; this is a different person from the subject of this page, and the sheet supplies the distinction itself. Read at the image at 400 dpi; the sheet’s own production stamp reads EFTA00701367. https://epstein-data.com/EFTA00701362 p.6 ↩

67. The remainder of the 125 one-f pages were read in the corpus text layer only and were not rendered. The one page that matches a single roster marker and carries no roster entry is https://epstein-data.com/EFTA02686105 p.61, a long alphabetical list of names in one column with a column of organisations beside it, in which “ACTIVATE Inc.” and “Michael Wolf” stand in different columns. Among them are a professional-network contacts export whose entry for a Michael Wolf gives a different occupation, and two long alphabetical lists of names in which the one-f spelling stands among other Wolf and Wolff surnames. Verifying the spelling on any of these at the image would cost one render each; the claim they support here is only that the one-f population is mixed, which the roster sheet at ⁶⁶ establishes on its own. UNVERIFIED. ↩ ↩² ↩³

68. Three messages spelled with one f in the House Oversight production or in one EFTA copy survive elsewhere spelled with two: a memo about a publisher of “Fri, Mar 18, 2016 at 11:39 AM” (https://epstein-data.com/EFTA02468452 p.1 and https://epstein-data.com/EFTA00831392 p.1); a forward of 1 February 2019 (https://epstein-data.com/HOUSE_OVERSIGHT_023044 p.1, and see ⁵⁹); and a forward of fact-checking questions of 31 March 2015 (https://epstein-data.com/HOUSE_OVERSIGHT_026351 p.1 and https://epstein-data.com/EFTA00860835 p.1). Established by phrase search across both spellings and read in the corpus text layer; the spelling difference between copies was not checked at the image on every one of them, and is not the basis for any claim on this page beyond the rejection of the one-f form as a union member. UNVERIFIED. ↩ ↩²