Caleb Harper

How he appears in these documents	Writing to Jeffrey Epstein in his own voice under the sender name “Caleb Harper”. The sign-offs the sheets actually print, each on two lines, are “Best” over “C”, “Best” over “Caleb”, “Best,” over “Caleb”, “Thx” over “C”, and a bare “C”123
Described in the released files as	”Principal Investigator / Director / Open Agriculture (OpenAG) Initiative / MIT Media Lab”, in his own signature block of 8 May 20174
Named in the release as	”Caleb Harper, the plant guy?” — the phrase is Joi Ito’s, written to Epstein’s assistant on 25 April 2018 when Ito was proposing who else Epstein should meet. Epstein’s office adopted it: a scheduling sheet lists “Caleb Harper (the plant guy)” and a subject line reads “Caleb Harper- the plant guy”. The label is not unique to him: in the same correspondence it attaches to Peter Wohlleben, to his own subordinate John De La Parra, and to at least one person the release does not identify56789
Span of the record	6 May 2017 to 22 May 2019, measured over 192 documents101112
What he asked Epstein for	18 June 2018: “I’d propose 500k a yr for 3 yrs”, itemised to three salary lines and a prototyping budget. Years 2 and 3 are conditional in his own wording, “with the expectation we evaluate progress after year one”. 19 October 2018: “the 500k for s[u]pport during dec 2018-2019 for the sentience work” — on the reading argued below, the first of those three years131
What Epstein wrote back	17 June 2018, the day before the itemised budget was sent: “ill fund it, mancuso also fun, .” 18 June, on seeing the itemised budget: “not of of the ballpark , can talk later in the weekm”. 5 July: “im ready”. 3 August, asked to fund a first hire: “yes,”. Harper’s own reply an hour later says what he then did with it: “I’m going to start with him on a shor[t] contract (6-8 weeks)“14131516
The route he described	Through MIT and through Ito — “I believe Joi has worked out the pipeline with MIT”; “Im working with Joi now to accept the funding”; “Ive cleared the ability to take the funds with Joi”11718
What the release records Epstein answering	8 November 2018: “Im waiting on joi”. 2 December 2018: “How much does john need assp”. 4 December 2018: “what do you need immediatly to keep john happy”192021
Whether any money reached him	No document read for this page records a payment to him, to his group or to the project. The release holds the request, the itemisation, the chasing and Epstein’s questions, and stops there22
Meetings arranged with Epstein	6 May 2018 in Cambridge, where a schedule lists him as confirmed; and Sunday 2 December 2018 at 2pm at 1 Brattle Square, Suite 6, with John De La Parra, which Harper accepted with “I’ll be there!”. A third, on 15 October 2018, he could not attend and sent De La Parra in his place62324
Not to be confused with	Charles L. Harper Jr., a different man of almost identical corpus size, whose documents do not intersect these at all25

Caleb Harper is a researcher who was at the MIT Media Lab. Emails released by the Justice Department under the Epstein Files Transparency Act record two years of contact with Jeffrey Epstein, from May 2017 to May 2019, in which Harper proposed a research programme on plant sentience, set out a budget for it of half a million dollars a year for three years, asked Epstein in writing whether they were “good to move forward with the 500k”, chased him when no answer came, and described the route the money would take as running through MIT and through the Media Lab’s director, Joi Ito.^111311812

This page is about what Harper himself did: what he asked for, when, through whom, and what the release records by way of an answer. The Media Lab’s wider funding relationship with Epstein is not restated here; it belongs to joi-ito and mit-media-lab, and the meetings held in Martin Nowak’s office belong also to martin-nowak.

A request for money is not a payment. Nothing on this page should be read as recording that Epstein gave Harper anything. What the documents establish is an ask, its size, its itemisation, its route, and a sequence of unanswered follow-ups.

He is a living person and is named here in full. No response by him to any of these documents is recorded in the release, and none is reproduced on this page.

Two things to get right before reading further

First, this is not Charles Harper. The release holds a second, unrelated Harper of almost exactly the same size. Charles L. Harper Jr., formerly of the John Templeton Foundation, appears in a 198-document name union whose traffic is overwhelmingly 2010. Caleb Harper appears in 192. No document is in both sets. A bare search for "Harper" returns 1,099 pages in 834 documents, of which 444 documents belong to neither man, and no claim on this page rests on one.²⁵

Second, some of these sheets carry transmission damage in the print itself. These are Gmail-export sheets on which a quoted-printable soft break has been carried onto the face of the document, so that an = stands where a character was lost, one character at a roughly fixed interval. It is what the document prints, not an artefact of the corpus text layer, and it has not been silently repaired here. Where a word is quoted with a character missing, the supplied letter is given in square brackets and the damaged form is shown as printed. Where the same message survives in an undamaged copy, that copy is cited beside the damaged one and named as the basis for the reading.²⁶

How the contact began, May 2017

The earliest documents in the run are not from Harper and are not about a research project. On Saturday 6 May 2017, at 18:23 UTC, Epstein wrote to Joi Ito. The whole of that message, as the sheet prints it, is:

Caleb email? =aby?

Ito answered seventeen minutes later, at 18:39:48 UTC. His reply opens with Harper’s name and an address the copy read here bars, and goes on: “Baby… counting down. Should be =n the next few days probably…”¹¹

The whole sequence takes twenty-five minutes. Epstein asked at 18:23 UTC; Ito answered at 18:39:48; and at 18:45, five minutes after receiving the address, Epstein sent a message whose entire body reads “Caleb Melz Melz Caleb”, a two-way introduction of a kind that recurs throughout his correspondence.²⁷ The other party wrote back three minutes after that: “Caleb! Hi - nice to be in touch - sounds from Jeffrey like a lot of great ideas in progress. Would love to talk more if you have a moment”.²⁷²⁸ Harper replied on Monday 8 May at 3:01 p.m. His message, in full, is “Hi”, then “Would be great to connect.”, then “JE - ok to move to BCC or want to stay in?”, then “C”, each on its own line.²⁹ Three minutes later he sent a second message adding, “+Whitney to help us coordinate schedules for a call/visit [a]t your convenience.”³⁰ The following day the third party wrote back to both of them: “Hi there to you both (JEE to bcc) … am working w”, at which point the line breaks off at a barred value and resumes, “on dates and when she finds one, back in touch … not sure JEE joining but hope so … Terrific to connect”.³¹

What that sequence establishes is narrow and worth stating precisely: Epstein asked Ito for Harper’s contact details, Ito supplied them, Epstein used them to make an introduction, and Harper took the introduction up. The third party’s name and address are barred in the headers of the copies read for this page and are not supplied here.

Harper’s signature block on his own message of 8 May 2017 reads “Principal Investigator / Director / Open Agriculture (OpenAG) Initiative / MIT Media Lab”, followed by “*Please contact Ms.Whitney Burke for scheduling”, and the sheet carries two links beneath it, one to his TED talk and one to a community forum.⁴

Four months later, on 7 September 2017, Ito forwarded Epstein a technical description of a gas chromatograph-mass spectrometer being bought for the lab, with a covering line of his own: “This machine will kick off the collaboration between Caleb and Ed to work on the plant/animal stuff.”³² The forwarded message beneath it is between three other researchers and Harper, and describes the instrument at length; it makes no mention of Epstein. The line is Ito’s characterisation of what the machine was for, and is reported here as that rather than as a fact about what Harper and the others intended.

April to May 2018: “Caleb Harper, the plant guy?”

The phrase that follows Harper through the rest of the release originates with Ito, not with Epstein. On Wednesday 25 April 2018, Epstein’s assistant Lesley Groff had written to Ito, “Great! …so far I have Martin, Joscha, Larry Summers…working on Mike Hopkins…Danny Hillis and George Church are not available…is there anyone you wanted him to meet with ?” Ito replied at 20:34 UTC:

Should I invite Madars as usual? Do you think he’d be interested in Gary Gensler? I’ll think of others. Caleb Harper, the plant guy? Ed Boyden?

The visit under discussion had been fixed earlier in the same thread: “Hi Joi…may we confirm 1-3pm on May 6th at Martin’s Institute..?”⁵ Groff forwarded Ito’s list to Epstein the same day with the note “below from Joi Ito…please advise”.³³

Ito used the label again six days later, and this is the sheet a name search does not reach, because the extraction renders “plant” as “plan”. On 1 May 2018 he wrote: “So far, I have Caleb the plan guy, Madars and Adam the new fund guy I hired coming.”³⁴

Epstein’s office adopted the label and kept it. A day sheet for the visit lists, under a block of time with Ito, “Confirmed Caleb Harper (the plant guy)”.⁶ In November a subject line reads “Caleb Harper- the plant guy”.⁷ In May 2019 Groff used it again in the course of passing his contact details to a third party.¹²

”the plant guy” is not one man, and the count says so

"the plant guy" returns 50 pages in 50 documents, and two damaged forms found by ocr_variants add two more: "the plan guy" and "the plant uy", one page each. Union 52, all opened. The decomposition matters more than the total. It classifies by what else the same page names, which is not always what the phrase refers to. Three of the 44 are Nowak’s reply of 27 November, where the phrase names John De La Parra and Caleb Harper is merely the other name on the sheet:

what the same page also names	documents
Caleb Harper	44
Peter Wohlleben	3
neither	5

The three are a different plant guy entirely. On 31 July 2018 Lesley Groff wrote to Epstein under the subject line “Reminder Peter Wohlleben the plant guy from NYBG Symposium”: “Would you still like me to try and coordinate a call for you with Peter Wohl[l]eben the Plant guy from the NY Botanical Gardens symposium.?” Epstein answered “yes”. A fortnight later Groff reported that she would “follow up on this again mid August…his publisher, who I have been coordinating with, is on holiday until Aug. 7.”³⁵

Of the five that name neither, two are a pair from 23 March 2018, a month before Ito first used the phrase of Harper. Epstein wrote to Groff, in a message whose other clause is about a different matter: “track down also lets get the plant guy from yesterday on the phone.”, and a calendar alert was raised the same morning reading “Track down the plant guy from yesterday and arrange a call”. No document read for this page identifies who that was.³⁶

Two more are copies of one message, and it is the sharpest document in this part of the record. At 13:45 UTC on 27 November 2018, replying to Martin Nowak’s question of 8:42 a.m. (“is Jeffrey coming on Sunday? (he mentioned the possibility)”), Lesley Groff wrote:

Good morning!! YES! He just told me. I was about to reach out to you. He says wheels up will be 10am. He would arrive Bedford around 10:30/10:45am. He wants to see the ‘plant guy’ he saw last time. Do you know who this is? Can you help?

³⁷

Harvard answered three minutes later. At 13:48:25 UTC Martin Nowak replied. His message carries an empty subject field and no quoted block, and its whole body is three lines:

this is the plant guy

please also try Caleb Harper

John De La Parra (from Caleb Harper’s group)

The third line is indented, set in italic type, and separated from the second by a double gap; a barred value follows it. It reproduces, word for word, the naming portion of the line that appears on Epstein’s own 15 October itinerary, “2:00pm Appt w/John De La Parra (from Caleb Harper’s group)”, without its time prefix. Groff answered at 13:49:19 UTC: “Ah ! Excellent! Thanks”.⁹ Twenty-two minutes after Nowak’s reply, the invitation went out to Caleb Harper.³⁸

The question Groff asked settles this, and the answer’s own grammar does not. She asked about the plant guy Epstein “saw last time”, and Epstein had already written down, in his own words, which man he had just met and what he thought of him.

On Saturday 20 October 2018, the day after Harper asked “What did you think of John?”, Epstein sent two messages to a third party from his own handset. The log timestamps them in UTC and the sheet prints them twice:

I just left the herbarium at Harvard. GREAT FUN!

Great new kid. John de la para . Super

³⁹

That is Epstein’s assessment of John De La Parra, volunteered to someone else, five weeks before Groff asked Nowak who “the plant guy” was. Harper is not mentioned. The spelling drops the second r, which is why no search on his name returns it.

The scheduling record agrees. Harper did not attend the 15 October Cambridge visit: he was in Scotland and sent De La Parra in his place, and Epstein met him there.²⁴¹ So on either reading of “last time”, the 15 October meeting or the herbarium on the 20th, the plant person Epstein had recently seen was De La Parra, whoever Nowak’s “this” was pointing at.

Three further things sit with that reading rather than against it. Twenty-two minutes after Nowak’s reply the invitation went to Harper and asked, in the same breath, about “your colleague John”.³⁸ The 2 December itinerary then carries both men at 2 p.m.⁷ And in the whole release an appointment in Harper’s own name resolves to exactly one day, 2 December 2018, on thirteen copies of one itinerary; he is listed for 6 May, but inside Ito’s block rather than as an appointment of his own.⁴⁰

The layout invites a reading this page does not rest on. Nowak’s third line is set apart and names De La Parra, so “this is the plant guy” can be read as pointing forward to a pasted item. That reading is available but not compelled: the next name after “this” is Caleb Harper; the gap separates the pasted line from both preceding lines; and an empty subject with no quoted block is equally the shape of a reply to an out-of-band message, Groff having just written “He just told me”. The conclusion above does not depend on it.

So on the one occasion the record lets us test the label, it fits Harper’s subordinate rather than Harper, with Harper named beside him. That is why this page treats “the plant guy” as an identification nowhere it is not accompanied by the name.

The fifth is on joi-ito, and that is why the document is left there. On 19 February 2019 Ito wrote to Epstein, “I got $467K from ASU. Thanks!! We’ll fund the plant guy.”⁴¹ It names no one. Whether it means Caleb Harper is an open question, and this page declines to answer it: the label demonstrably attaches, in this same correspondence, to three identified men (Harper, Peter Wohlleben and John De La Parra) and to at least one the release does not identify, and nothing read here settles which is meant. Even read in Harper’s favour it would record an intention, not a payment.

The visit itself was Sunday 6 May 2018. The day sheets put Martin Nowak at 10:15 a.m. and a long block with Ito from 11:30 a.m. to 3 p.m., inside which the confirmed names are “Madars”, “Adam (the new guy Joi hired)”, “Caleb Harper (the plant guy)” and a plant scientist whose forename the sheets render three different ways. Joscha Bach takes 3 p.m. and a dinner with Larry Summers closes the day; a later copy of the same schedule adds Leonard Peshkin at 4 p.m.⁶ A schedule records an appointment, not an act, and no document read for this page describes what was said at that meeting.

What is recorded is that at 17:44 UTC on that Sunday, Harper sent Epstein a message with the subject “Book on sentient plants”, whose entire body is the word “Semiosis” and a link to a novel of that name.⁴² It is the earliest document in the run in which Harper writes to Epstein about the subject that the rest of the correspondence turns on.

The ask, June to October 2018

17 June 2018: “ill fund it”

The funding conversation opens on a Sunday evening. At 2:31 p.m. on 17 June 2018 Harper sent Epstein a message whose body cannot be read. It was an inline image: the sheets that carry this thread print the message’s header and then nothing, and one of them prints, at exactly that position, an image tag in place of a body. The sheet that carries Epstein’s replies names the attachment in its own header. The production does not render it, and this page therefore makes no claim about what that message said.⁴³ Epstein answered at 5:57 p.m.: “fun isnt it”. Harper replied at 6:08 p.m.: “Very…I wonder if I can prove it”, followed by a thinking-face emoji. Epstein’s answer, at 6:10 p.m., is in full:

ill fund it, mancuso also fun, .

That line is read here at the image on an undamaged copy; on a second copy of the same thread the extraction drops its body altogether.¹⁴⁴⁴ “mancuso” is not further identified by any document read for this page.

Two things follow from that message and no more than two. Epstein wrote that he would fund the work, in his own words, the day before the itemised proposal arrived. And nothing on this page establishes that he did.

A narrower statement about the figure can be made, and it is the one this page makes. No sum appears anywhere in the 192 documents earlier than the budget of the following morning: "500k" occurs in 24 of the 192, every one of them a copy of the 18 June thread or of the October-November “the 500k” thread, and "500,000", "1.5 million" and "1,500,000" occur in none of the 192. That is a statement about what the release holds, not about what the unreadable message contained, and the distinction is the point.⁴⁵

18 June 2018: the itemised budget

At 11:32 a.m. the next morning, Monday 18 June 2018, Harper sent the request in full. It begins “I’ve ordered it :) thx”, a reply to something else in the thread, and continues:

To do this id need to add depth to my bio team, that’s by far my weakest team in the lab currently. I recently let contracts lapse on Arielle and my molecular biologist as it wasn’t working out. I’d want to add the following team and would need to do a search to get the best. MIT won’t let me hire PhD Research Scientists, and likely quality people wouldn’t accept, without secure multiyear funding. I’d propose 500k a yr for 3 yrs with the expectation we evaluate progress after year one as to year 2 and 3. The transparent cost break down is:

PhD plant physiologist - 80k base salary +56% mit research OH (none goes to my lab or ML) + 20% benefits = 149, 760 k total cost

PhD plant biochemist - 100k base salary + 56% mit research OH + 20% benefits = 187k total cost

Plant science tech (JR) - 40k * 20% administrative OH - 48k total cost

100k prototyping budget + 20% material & services OH = 120k total cost

= 500k a year / 3yrs

It goes on to state what Harper was putting in against it: “On my side I’ll dedicate 2 out of our 4 shipping containers for 3 years (target paid for install / dev etc but with stock price decline is stopping support for non essential research) and my salary is covered in other sponsored accounts plus my team of engineers on staff”, and closes: “I know it’s a big ask, and would understand if it’s out of ball park, but I think that’s what it would take to make meaningful and rigorous progress on figuring out if plants can think and communicate”, followed on a new line by “Thoughts?”¹³

That text is quoted from an undamaged copy. A second copy carries the same passage with transmission damage, and a third re-sends it in September under the heading “Previous[ ]email breaking down funding”; the wording is the same on all three.⁴⁶⁴⁷

The request is therefore for up to $1.5 million, not for $500,000, with the second and third years conditional in Harper’s own wording: “with the expectation we evaluate progress after year one as to year 2 and 3”. The figure that recurs later, “the 500k”, is one year of it.

Three features of the request are worth separating from the number.

• It is a request for salaries at MIT’s own overhead rates. Harper notes of the 56 per cent research overhead that “none goes to my lab or ML”.
• It is a request for multiyear money, justified by an institutional constraint: “MIT won’t let me hire PhD Research Scientists, and likely quality people wouldn’t accept, without secure multiyear funding.”
• The four line items sum to $504,960, which is the arithmetic behind “500k a year”: 80,000 raised by 20 per cent and then by 56 per cent is 149,760 exactly, and the other three lines compute the same way. The stated total is the document’s own and has been checked rather than taken on trust.⁴⁸

Epstein replied eight minutes later, at 11:40 a.m.:

not of of the ballpark , can talk later in the weekm, what happened to the two you let go

Harper answered at 11:52 a.m. that one researcher “wanted to get out of the lab and move back to NYC with her husband” and that the second “just wasn’t working out in terms of progress”.¹³

5 July 2018: “Checking back in on this”

Nothing further in the thread is recorded for seventeen days. On Thursday 5 July 2018 at 11:18 a.m. Harper wrote, in full:

Checking back in on this

Thx C

Epstein replied at 15:24 UTC:

im ready . can discuss more when i return to states next wed great

¹³

This is the first of four occasions on the record on which Harper chases an unanswered funding question, and the pattern set here holds for all of them: a short prompt from Harper, a short and non-committal reply from Epstein.

1 August 2018: “I’d like you to fund”

On Wednesday 1 August 2018 at 17:04 UTC Harper wrote under the subject “Re: Scientist”, opening with a link to a professional profile and continuing:

I[‘ll] send you some of his papers soon. He seems like an ideal candid[a]te to join the team and is looking for a research scientist position curre[n]tly. Had the first interview and he passed with flying colors in term of t[e]chnical depth, experience and attitude.

This wou[l]d be the first guy to join the team researching plant sentience I’d like you to fund if you are still interested.

He[ ]is coming back next Thursday for more in-depth brainstorming and design of[ ]experiment review with us.

What do you think?

The man in the link is John De La Parra.⁴⁹ Epstein’s reply, sent at 01:17 UTC on Friday 3 August, is one word and a comma:

yes,

Harper answered sixty-two minutes later, and his answer is what “yes,” actually produced. Both stamps are the same unlabelled Sent: field on two sheets of one export, so the interval holds whatever the zone, which matters because neither message carries an RFC header:

Sweet! I’m going to start with him on a shor[t] contract (6-8 weeks) to make sure he is the right fit.

The limit is Harper’s, not Epstein’s. Epstein wrote one word; it was Harper who decided, an hour later, to start De La Parra on a short contract, and he says so in the first person. Nothing on the sheet records Epstein setting a term, a sum or a condition. The full-time appointment was still being negotiated five weeks later, when Harper reported De La Parra “on board as a temp to perm hire” and willing to go full time “starting in Nov”.¹⁵¹⁶⁴⁷

The sentence quoted above survives in three copies and no one of them is clean throughout; each carries the transmission damage at a different point, and the reading here is composed from all three.⁴⁹¹⁵¹⁶

The same thread carries, further down, a message Epstein had forwarded to Harper on 17 July from the biologist Merlin Sheldrake, reading “Great to chat, exciting offer. Am getting thinking”, and listing three papers on plant signalling.⁵⁰

6 and 24 September 2018: the route, stated three times

On 31 August 2018 Epstein sent Harper a message of two words: “nexzt step?”⁵¹ Harper’s answer, on 6 September, is a numbered list. Item 1 reports that De La Parra was “on board as a temp to perm hire (he is [p]erfect for the team - brilliant)”, “willing/able to leave his positions at North Eastern [a]nd Tufts and go full time with us starting in Nov”, and until then “committed to 50% time”. Item 2 names Ed Boyden, Martin Nowak and the network scientist Albert-László Barabási as collaborators, proposes developing a quantified “turing test” for plants as a first step “to proving sentience”, and asks whether Epstein knows “any[o]ne that could collaborate” as “a reputable thinker on consciousness, maybe a philosopher?” Item 3 is headed “Funding”:

I cant hire John full time and start spending money on other staff/proto[t]ype development/experiment costs until Joi signals to me that the budget w[e] talked about has landed at ML. Is it possible to move forward with that?

Item 4 attaches two papers, one of them “a whit[e] paper John started in the last week or so geared towards using our tools [f]or drug discovery”.⁴⁷

On 24 September 2018 Harper sent a message headed “Project Next Steps” with two attachments. The first is named on the sheet as 1_Page_Summary_Plant_Sentience.docx; the second attachment’s filename is barred but for its ending, (DRAFT).docx. The message reads:

TLDR: Im working with Joi now to accept the funding and he needs to know when it would start and how (from whom?) we would receive it to move forward with logistics. I would prefer a Dec. 01 2018 start date if possible. I have the plant biochemist I mentioned to you set up to start then and that would give me time to identify a proper bio-statistician. IF thats OK I’ll move forward with that.

He described one attachment as a “draft outline doc for internal use for Joi to use to coordinate funding with VP office, etc.”¹⁷ Epstein’s reply the next day, in full, is “look at pollen grains and virus structures?”, to which Harper answered “Hmmm interesting idea -”.⁵²

“How (from whom?) we would receive it” is the sharpest line in the run, and it is Harper’s own: three months after Epstein wrote “ill fund it”, the source of the money was still not something Harper could state.

The proposal itself

One of the two attachments survives in the release as a document of its own, and it is not in the population that a search for his name returns, because its byline abbreviates his forename. EFTA00788080 is a one-sheet paper headed “Plant Sentience”, bylined “C.Harper” and “J. de la Parra” and dated “9.22.18”, two days before the message that attached it. Its bold emphasis is the document’s own:

Testing and proving advanced philosophical concepts of plant sentience will rely first on the development of a nuanced lexicon for plant communication. At MIT’s OpenAg Initiative we have a rare set of tools, unique thinkers in plant science, and unparalleled engineering skills that will allow us to develop a comprehensive inventory of plant signaling — a plant dictionary for human understanding.

It defines sentience as “the ability to perceive stimuli”, lists a “conservative list of plant ‘senses’” running to nine named mechanisms and notes that “by some accounts there are up to 15”, and describes the method as the “construction of rigorously controlled environments” in which isolated stimuli are applied and every parameter recorded, in order to characterise “the resultant plant ‘chatter’, ‘singing’, ‘crying’, ‘warning calls’”.⁵³

This is the document the $500,000 was for, and it is the only sustained statement of the work in Harper’s own hand in the release. A search for "Caleb Harper" does not return it; "plant dictionary" returns it and nothing else.

”Are we good to move forward with the 500k”

On Friday 19 October 2018, at 18:22:52 UTC, Harper sent Epstein a message under the subject “Moving forward”. It is the single document on which this page principally rests, and it survives in at least four copies whose transmission damage falls in different places. On one copy the sheet prints:

What did you think of John?

=br>

Are we good to move forward with the 500k for s=pport during dec 2018-2019 for the sentience work?

I believe Joi has worked out the pipeline=with MIT but time is of the essence to secure John to the team

Thx

C

Three of the marks on that sheet are transmission damage carried onto the face of the print, not the writer’s own text. The reading of the two damaged words is not guessed: a second copy of the same message, in a different typeface and with no damage at either point, prints “for support” and “the pipeline with MIT”, and a third copy, quoting the message inside a later reply, prints both words undamaged as well.¹⁵⁴¹⁸ The line reading “=br>” is a fragment of an HTML line-break tag from the message’s own markup; it stands where the paragraph break is on the clean copy.

So the undamaged text of the request is:

What did you think of John?

Are we good to move forward with the 500k for support during dec 2018-2019 for the sentience work?

I believe Joi has worked out the pipeline with MIT but time is of the essence to secure John to the team

Thx C

This is a named researcher, in his own voice, asking Jeffrey Epstein in writing for a specified sum, over a specified period, for named work, and naming the director of the MIT Media Lab as the person who had arranged the route. It is one sheet, and it needs no inference to read.

Note what the request is and is not. The “500k” here is the first year of the three-year proposal of 18 June, not the whole of it: the period named is “dec 2018-2019”, and the start date Harper had asked for in September was 1 December 2018.¹³¹⁷ It is also not the $500,000 that reached Joi Ito’s discretionary account in May 2014. That was a different sum, from a different source, four years earlier, into an account Ito and Epstein both called his “slush fund”; it is described on joi-ito and has nothing to do with this request beyond the coincidence of the figure. The two must not be run together.⁵⁵

Twenty days of silence, then “Im waiting on joi”

No reply to the 19 October message is recorded. On Thursday 8 November 2018 at 13:33 UTC Harper wrote again:

Checking in again. Don’t mean to be a nuisance but moving forward on hiring John for the plant sentience project and experiments hinges on funding. If you are still interested in supporting the project it would be really helpful to know. Ive cleared the ability to take the funds with Joi.

Best C

Epstein replied thirty-two minutes later, at 14:05 UTC. His answer, in full, is:

Im waiting on joi

Both are read here at the image. “Ive” is as printed, without an apostrophe, on a copy that carries the whole message undamaged, and every text layer of every copy reads it as “lye”, which is the extraction and not the sheet. “Im” is likewise as printed.¹⁸¹⁹

Read against the 19 October message, that reply inverts the account Harper had been giving. Harper had told Epstein three times that Ito had the arrangements in hand: “Joi has worked out the pipeline with MIT”, “Im working with Joi now to accept the funding”, “Ive cleared the ability to take the funds with Joi”. Epstein’s answer says the hold-up is Ito. Which of the two descriptions is right is not settled by anything on this page, and no document read here records Ito being asked.

December 2018: two questions, no answer recorded

The next documents in the funding thread are both Epstein’s, and both are questions rather than answers.

On Sunday 2 December 2018 at 23:49 UTC, the evening of the day Harper and De La Parra were scheduled to meet him in Cambridge, Epstein wrote to Harper. The message carries no subject line and its entire body is:

How much does john need assp

Two days later, on Tuesday 4 December at 10:40 UTC, he wrote again. That message is also one line:

your gmail is not working,. what do you need immediatly to keep john happy

Both are quoted as printed; “assp”, “immediatly” and the stray comma-full stop are the sheets’ own.²⁰²¹ The remark about the mailbox is corroborated twice over. On 28 November Lesley Groff had written to Harper, “Hi Caleb. I just received a bounce back email stating the mail has been delayed for some reason. Just want to make sure this is getting through to you.”²³ And on 29 November Groff forwarded a delivery-failure notice to John De La Parra with the line “This is what I keep receiving…”. It reports that the invitation of 27 November had not been delivered, the account being “over quota”, and it carries that invitation at its foot: “Hello Caleb. Jeffrey Will be up at Harvard this coming Sunday Dec 2. Might you be around and available to meet with him? He was also wondering about your colleague John?”³⁸

Those two messages are the last documents in the release, on any copy read for this page, in which Epstein and Harper discuss money. No reply by Harper to either is recorded here.

Did any money move?

Epstein wrote that he would fund the work. No document read for this page records that he did. Those are two different statements and the page keeps them apart.

On 17 June 2018 Epstein wrote “ill fund it”. On 3 August, asked to fund a specific hire, he wrote “yes,”, after which Harper said he would start the man on a six-to-eight-week contract. On 5 July he wrote “im ready”. Every one of those is an expression of intent, in a private message, with no sum agreed and no route settled in any document read here, and on 8 November, four months after “im ready”, Harper was still writing that “moving forward on hiring John for the plant sentience project and experiments hinges on funding”, and Epstein was still answering “Im waiting on joi”.

Against that: no document read for this page records a payment, transfer, wire, cheque, invoice, gift letter or receipt to Caleb Harper, to his group, or to the plant-sentience project, and none records anyone saying that one was made.

The searches behind that statement, so a reader can re-run them. Returning 0 pages: "Caleb Harper" with "wire"; with "500"; "Harper" with "wire transfer"; "gift information" with "Harper"; "sentience" with "gift"; "De La Parra" with "500". "Caleb" with "wire" returns 3 pages in 3 documents and "Caleb" with "payment" 10 pages in 10 documents; the two sets share one document, so the union is 13 pages in 12 documents, and all 13 pages were opened and read. Not one concerns this Caleb: they are a federal courthouse named for Caleb Boggs, an immigration official, a Miami courthouse named for Joseph Caleb, a concrete-pumping contractor named Uriel Caleb, two copies of a Bloomberg article by a reporter named Caleb Melby, and two extractions in which “Caleb” is scanner noise.⁵⁶

What a disconfirming document would say, in its words rather than in mine. It would report that the funds “have landed at ML”, which is Harper’s own formula for the event he was waiting for; or it would be a gift or wire confirmation naming an amount of 500,000 or 505,000 alongside MIT or the Media Lab; or it would be a message from Harper acknowledging receipt; or it would be an MIT gift letter of the kind this release holds for other gifts. "landed at ML" returns 4 pages in 4 documents, all four of them copies of Harper’s own message of 6 September asking whether it had happened; all four were opened. "has landed" returns 191 pages in 189 documents, which is too large a set to read whole and was not read whole; it is closed below by conjunction rather than by reading.⁵⁷

The strongest test is not a phrase search at all: it is a vocabulary sweep over the whole population. Twenty money words were run across every sheet of all 192 documents, not over a sample and not over a classified subset:

• wire, invoice, cheque, deposit, remit, disburse, payment, transfer, donate, gift, grant, reimburse, honorarium, stipend: zero occurrences anywhere in the 192.
• paid occurs 17 times and is one sentence, “target paid for install / dev etc”, from the 18 June budget, rendered with different transmission damage across copies.
• money occurs 4 times, all the 6 September funding paragraph. funds 4 times, all “Ive cleared the ability to take the funds with Joi”. landed 4 times, all “has landed at ML”. account 17 times, all “my salary is covered in other sponsored accounts”.
• receive and its forms occur 142 times across 85 documents, and every occurrence was classified: 119 are the confidentiality footer, 12 are Gmail date-received metadata, 9 are Harper’s own forward-looking question of 24 September (“how (from whom?) we would receive it”), and 2 are Lesley Groff reporting the bounced invitation. The residual is 23, and none of the 23 is a receipt of money.

So every money-vocabulary occurrence in the entire population is a copy of a message already quoted on this page. That result does not depend on the header classifier, on which documents were counted as correspondence, or on any sampling judgement.²²

And the one unread count can be closed. "has landed" returns 191 pages, too many to read whole; conjoined, "has landed" AND "media lab" returns 0, AND "MIT" returns 1 (EFTA02619217, already read here) and AND "Harper" or AND "Caleb" returns 4, the same four copies of Harper’s own question. "De La Parra" conjoined with "salary", "hired" or "paid" returns 0.⁵⁷

So the claim this page makes is bounded and falsifiable: within the 192 documents that name him and the 13 further pages read from the “Caleb” money searches, the release records the asking and not the paying. It does not establish that nothing was ever paid: a payment could sit under a name other than his, in a production not searched here, or nowhere in this release at all. It establishes that nothing read for this page shows one.

One further document exists and is deliberately left where it is. On 19 February 2019, eleven weeks after the last of these messages, Ito told Epstein that $467,000 had arrived from Arizona State University and added, “We’ll fund the plant guy.”⁴¹ That sentence is on joi-ito, where it belongs, and it names no one. It is treated above as an open question rather than a finding, for the measured reason that “the plant guy” in this correspondence is at least three different people. Even taken at its strongest it records an intention to fund, not a payment.

The meetings, and the man he sent in his place

Three Cambridge visits by Epstein touch Harper in the record. He was scheduled for two of them. (On 20 October 2018 Epstein’s own handset places him at the Harvard herbarium while his office’s day sheet places him in Palm Beach;⁵⁸ the conflict is recorded rather than resolved, and no document read for this page connects Harper to that day either way.) Only the arrangements are documented; no document read for this page describes what was said at any of them, or confirms that any of them took place.

Sunday 6 May 2018, Cambridge. Arranged through Lesley Groff and Ito over a fortnight of messages and confirmed as “1-3pm on May 6th at Martin’s Institute”, with Harper listed as confirmed inside Ito’s block; the day sheets are set out above.⁶ At 17:44 UTC that day Harper sent the “Book on sentient plants” message.⁴²

Monday 15 October 2018, 1 Brattle Square. On 6 October Epstein had written to Ito and Nowak, “any day good week of the 15th for a visit. caleb ? etc.”⁵⁹ On Thursday 11 October at 5:24 p.m. Harper answered the invitation:

Joi forwarded me a message about mtg w/ JE on Monday. Unfortunately I am in Scotland but I would suggest that he meet with John De La Parra from my group and I have checked that he can meet between 1-5p. He is CCd here for your convenience.

Lesley Groff replied forty-one minutes later: “Hello Caleb and John. Jeffrey would love to meet with John on Monday Oct. 15th…John, could you please come to Martin Nowak’s office at 1 Brattle Square, Suite 6 at 2pm?. Sorry we will miss you Caleb… we will try again next time!” De La Parra answered three minutes after that, “Absolutely. I will be there”, and Groff closed it: “Fantastic! confirmed.”²⁴ This is the meeting Epstein was recalling five weeks later when he asked to see “the plant guy he saw last time”, and Harper was not at it.

Sunday 2 December 2018, 1 Brattle Square. On 27 November Groff wrote asking whether Harper and De La Parra could meet Epstein at Nowak’s office on 2 December. Harper answered at 4:40 p.m.: “I’ll be there!” The confirmation, at 23:29 UTC the same day, reads: “Hello Caleb and nice to hear from you also…so you and John are confirmed to meet with Jeffrey on Sunday Dec. 2nd at 2pm in Martin Nowak’s office- 1 Brattle Square, Suite 6!”²³⁶⁰ The itinerary for the day, circulated under the subject “Caleb Harper- the plant guy”, carries Nowak at 11 a.m., De La Parra and Harper both at 2 p.m., Danny Hillis at 3 and a tentative George Church at 4.⁷ On the morning of the meeting De La Parra confirmed once more: “Great thank you! We will be there at 2.”⁶¹ It is the only day on which Epstein’s schedules give Caleb Harper an appointment in his own name.⁴⁰

1 Brattle Square, Suite 6, is Martin Nowak’s office at Harvard, and it is named as such on the sheets: a place of business, and the venue named for every Cambridge meeting in this run.

John De La Parra

De La Parra is the person the money was for, and everything said about him above is his: the candidate of 1 August, the one-word “yes,” of 3 August, the temp-to-perm hire of 6 September, the man sent to Cambridge on 15 October in Harper’s place, and the “john” of both Epstein’s December questions. Epstein’s own schedules call him “from Caleb Harper’s group”,⁶² and it is his name, not Harper’s, that Epstein wrote down the day after the October visit: “Great new kid. John de la para . Super”.³⁹ He is also, on the reading argued above, the likelier referent of “the plant guy” in Nowak’s reply of 27 November; that reading is offered with its limits stated rather than as settled.⁹

He writes in his own voice on several of these sheets, and his name returns 85 documents in its full form and 88 on the surname alone.⁶³ He belongs here rather than on a page of his own. Measured on his own population rather than on his prominence in this one, about half of those documents are copies of Epstein’s diary, and what he does in the rest is confirm meetings: “Absolutely. I will be there”, “Great thank you! We will be there at 2.” Being the person a funding request was built around, and being the likelier referent of someone else’s label, raises how much he matters here; neither is an act of his own.

After December 2018

The funding thread stops on 4 December 2018. The record of contact does not.

On 22 May 2019, five and a half months later and six weeks before Epstein’s arrest, Groff wrote to Victoria R. Montgomery Brown, chief executive and co-founder of Big Think: “Hello Victoria. Jeffrey requesting I send you the email address for Caleb Harper, the plant guy!” Montgomery Brown replied to Groff at 2:11 p.m., writing “Thank you very much! I am planning on following up with Jeffrey with some thoughts — should I email them directly to you or does he have an email address? Thank you so much for setting up the call. I’ll reach out to Caleb.”, and, after Groff answered “HI Victoria…you may certainly forward me any thoughts and I can forward to Jeffrey for you!”, closed the exchange at 19:06 UTC: “Wonderful! Thank you, Lesley.”¹²

That is the last document in the population read for this page. Every date on every sheet of all 192 documents was extracted and sorted: the latest is 22 May 2019, carried by three documents, all of them copies of this one exchange; the four further documents reached by the other search forms are all dated November or December 2018, and the “Plant Sentience” paper is dated September 2018. A disconfirming document would carry a date after 22 May 2019 beside his name; none does. It records Epstein still making introductions on Harper’s behalf five and a half months after the last message about money, and it is the reason this page’s span ends in May 2019 rather than December 2018.⁶⁴

What is not here

• "food computer" returns 0 documents, and "personal food computer" returns 0. The work Harper is best known for outside this release is not in it, under that name. "OpenAg" returns 19 documents and "Open Agriculture" 13, a union of 21; all 21 were opened. Twenty are his own signature block, a link to a community forum, or a scheduling line naming the initiative (“Caleb (OpenAg)”, “openAg head”) rather than describing it. The twenty-first is the “Plant Sentience” paper, which is quoted above.⁶⁵
• No statement by Harper responding to any of these documents is in the release, and none is quoted here from outside it.
• No document read for this page records anyone putting Epstein’s November answer to Ito, or records what MIT’s own gift office was told, if anything. Those questions belong to joi-ito and mit-media-lab.
• Nothing on this page is a finding of wrongdoing by anyone named on it.

How this page was built, and what it measures

The population. A full-index FTS5 phrase search of text_content for "Caleb Harper", run on 11 September 2026 and never scoped by dataset, returns 217 pages in 192 distinct documents: 190 under EFTA identifiers and 2 under HOUSE_OVERSIGHT_. scripts/ocr_variants.py "Caleb Harper" --count generated one measured-glyph variant, which returned zero; --wide generated 32 further single-edit forms, of which two returned anything: "Caleb Harpe" (5 pages, 5 documents, 1 outside the 192) and "Caleb arper" (1 page, already inside). Three further documents were reached by searching a header-field form rather than the name. The union of all forms is 196 documents, and two further documents reach him under forms no search on his full name finds: EFTA00788080, whose byline abbreviates his forename, and EFTA00472967, which calls him “Caleb the plant guy”. The stated figure of 192 is therefore a floor, and the page says so wherever it uses it.¹⁰

Act versus mention, re-derived over the whole population rather than a sample, and re-derived twice more after two defects in the method were found. Every one of the 192 documents was classified from the header block on its own first sheet (not on the first sheet that happens to carry his name), with whitespace flattened before any pattern was applied, and matching the sender label in its corrupted forms as well as its correct one:

	documents	share
Harper is the top-level sender	40	20.8%
addressed to him at the top level	35	18.2%
named by a third party	112	58.3%
no e-mail headers at all	5	2.6%

Direct two-way correspondence between Harper and Epstein: 75 documents, 39.1% of the population. All 75 were read in full.²²

Three faults in the method were found and fixed, and each moved the answer. Classifying on the first matched page rather than on page 1 left eleven documents’ headers unread and cost seven reclassifications, cutting the header-less class from 12 to 5. There are two header layouts, not one: 173 sheets set a label then its value, but 19 stack every label first and every value after, which makes a label-then-value parser assign the wrong value to every field; handling both moved three more. And 32 of the 192 print the sender label with nothing after it, of which 22 had fallen to “third party” on that account alone; all 22 were opened and one is a message Harper wrote, counted here as his.

The sender label is itself corrupted elsewhere in this release, and that was checked here rather than assumed. Across the 192, the label appears as From: in 177 documents and as Fran: in 2; Front, Frorn and Fronl appear in none, and neither of the two Fran: documents is one Harper wrote. So the corrupted-label defect is real in this population but changes no figure in the table above: a classifier matching only From: and a classifier matching the full label set returned identical results on all 192 documents.

The classifier’s residual limit runs one way. A document carrying no e-mail headers, or whose sender is barred and whose body is not signed, falls to “named by a third party” whatever it is. So the third-party share is an upper bound and the two direct shares are floors. The five header-less documents were each opened and are all day sheets or itineraries from Epstein’s office.

A second, independently written parser run against the same 192 gave higher direct counts than this one: 49 sender, 44 addressed, 94 third-party, 5 header-less. The header-less class agrees exactly, on the same five identifiers; the two direct classes do not. Neither parser is authoritative and the disagreement is not resolved here, which is why the table above is stated as floors, and why the money finding is based on a sweep of all 192 documents rather than on the direct subset.

And one authored document sits outside this table entirely. EFTA00788080, the “Plant Sentience” paper bylined “C.Harper”, is a document he wrote; it is not in the 192 because its byline abbreviates his forename. Counting it, the release holds 41 documents he authored.

What was read at the image, and what was not. Every quotation on this page from a message between Harper and Epstein was verified on a rendered sheet at 400 dpi, and where the wording was in doubt it was checked against a second copy of the same message in a different rendering. Forty sheets from 36 documents were rendered for this page, and 33 sheets from 31 documents were opened and read. A rendered sheet nobody opened is not evidence, so only those 31 documents are cited as image-verified, and every footnote that does not carry an image marker is read from the text layer. Two classes of statement were not image-verified and are marked as such: the corpus-wide counts in this section, which are properties of the index rather than of any sheet, and the contents of documents that are only counted here and not quoted.

Clock times, and a correction. This production carries three kinds of timestamp and they are not on one clock. An RFC Date: header states its offset and is quoted here as UTC. A bare Sent: field, one carrying no zone, is in UTC, and an explicitly zoned Date:/Sent: field is a different export shape that is usually Eastern: on two matched pairs of copies of one message, Sent: Saturday, May 6, 2017 6:40 PM sits against Date: Sat, 06 May 2017 18:39:48 +0000, and Sent: Friday, October 19, 2018 6:23 PM against Date: Fri, 19 Oct 2018 18:22:52 +0000. A quoted in-thread header (“On Fri, Oct 19, 2018 at 2:22 PM”) is local, four hours behind. An earlier version of this page stated the opposite rule for Sent: fields, and got one sequence wrong in consequence; that sequence is corrected above. Most intervals given on this page are computed within one kind of stamp. The 6 May 2017 sequence above is the exception and says so: it runs between an RFC header and quoted headers converted at four hours, and is offered on that basis.²⁸

Where a copy carries transmission damage and another does not, the undamaged copy is the one quoted, and the damaged one is named beside it. Nothing has been silently repaired; every supplied character is in square brackets.²⁶

Related pages

joi-ito · mit-media-lab · martin-nowak · jeffrey-epstein · charles-harper · epstein-files-transparency-act

The money finding in the body does not rest on that classification. It rests on a separate sweep of twenty money words over every sheet of all 192 documents, run without reference to the header classes: wire, invoice, cheque, deposit, remit, disburs*, payment, transfer, donat*, gift, grant, reimburs*, honorar* and stipend return zero occurrences anywhere in the 192. paid returns 17 occurrences in 17 documents and 10 distinct contexts, all of them one sentence rendered with different transmission damage. money returns 4 in 4, all one paragraph; funds 4 in 4, all one sentence; landed 4 in 4, all one sentence; account 17 in 17, all one clause. receiv* returns 142 in 85, classified exhaustively as 119 footer + 12 date-received + 9 “we would receive it” + 2 Groff on the bounced invitation = 142, a residual of 23 above the footer. An earlier count of this residual gave 21, because the classifier’s footer pattern also swallowed Groff’s two lines; the parts summed to 142 either way, which is why the error survived one pass. None of the 23 is a receipt. Every money-vocabulary occurrence in the population is a copy of a message quoted on this page, and none records a transfer, a receipt, or an acknowledgement of one.

Footnotes

1. Caleb Harper to Jeffrey Epstein, “Moving forward,” 19 October 2018, 18:22:52 UTC. Read at the image at 400 dpi; the quotation reproduces the damage as printed. https://epstein-data.com/EFTA02619312 p.1. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

2. The top message of the same sheet, 12 October 2018, which the sheet prints as “Sorry to miss you Monday - you’ll enjoy John=… interested to hear your thoughts after” — the = is transmission damage and is shown here as printed. Read at the image at 400 dpi. https://epstein-data.com/EFTA02619217 p.1. ↩

3. Every top-level sign-off in the 192 documents was extracted by pattern and counted; the pattern does not reach lines prefixed > or », and including quoted blocks the total is 47 rather than 28. The printed shapes, with the label on one line and the name on the next, are “Thx” over “C” (8), “Best” over “C” (7), “Best” over “Caleb” (7), “Best,” over “Caleb” (3), “best” over “C” (3), plus bare “C”. The string “Best, C” is printed by no document in the release: a literal test returns 0, and a test for “Best,” followed by a line beginning “C” also returns 0. An earlier version of this row quoted it, having welded the comma of one shape to the initial of another. A phrase search does not catch that error; it appears to confirm it, because the index drops the comma and matches across the line break, returning 122 pages for a string no sheet prints. Only a literal test against the text, or the image, settles it. ↩

4. Harper’s own signature block, 8 May 2017, on p.1; the two links sit on p.2, and p.1 ends at the scheduling line followed by a barred block. Read at the image at 400 dpi, p.1 only; p.2 is read from the text layer. https://epstein-data.com/EFTA00682099 pp.1–2. Later copies of the block substitute a different named scheduler: see https://epstein-data.com/EFTA01023259 p.3, not read at the image. ↩ ↩²

5. Joichi Ito to Lesley Groff, copied to Yuko Barnaby, “Re:”, 25 April 2018, 20:34:50 UTC, whose quoted block carries the whole scheduling thread back to 19 April 2018. Read at the image at 400 dpi. https://epstein-data.com/EFTA00472867 p.1. ↩ ↩²

6. Day sheets and schedules for the visit of 6 May 2018. Read at the image at 400 dpi: https://epstein-data.com/EFTA00473300 p.1 (Lesley Groff to Martin Nowak, 3 May 2018, “Below is the confirmed schedule so far”). Further copies, read in the text layer only: https://epstein-data.com/EFTA00285565 p.2; https://epstein-data.com/EFTA00285569 p.2; https://epstein-data.com/EFTA00307928 p.2. The plant scientist’s forename is rendered “Arielle”, “ArieHe” and “Arlene” across these copies. ↩ ↩² ↩³ ↩⁴ ↩⁵

7. Lesley Groff to Jeffrey Epstein, “Caleb Harper- the plant guy,” 27 November 2018, 11:33:21 PM, carrying the itinerary for 2 December 2018. Read at the image at 400 dpi. https://epstein-data.com/EFTA02262756 p.1. Several appointment lines on this sheet carry barred text after the name. ↩ ↩² ↩³ ↩⁴

8. "the plant guy" returns 50 pages in 50 documents. scripts/ocr_variants.py on the phrase produced 35 forms, of which two returned anything: "the plan guy" 1 page (EFTA02245244) and "the plant uy" 1 page (EFTA02462147), neither in the base set. Union 52 pages, all opened. Classified by what else the same page names: 44 name Caleb Harper, 3 name Peter Wohlleben or the NYBG symposium, 5 name neither. The five are two copies of the 27 November 2018 message at ³⁷, the two documents of 23 March 2018 at ³⁶, and EFTA01030310, Ito’s message of 19 February 2019. 44 + 3 + 5 = 52, and the three parts were enumerated by reading, not inferred from the total. ↩

9. Martin Nowak to Lesley Groff, 27 November 2018, 13:48:25 UTC, quoted here in full. Read at the image at 400 dpi: the subject field prints with an empty value, no message is quoted above the body, the third line is in italic type, and a barred value follows it on the same line. https://epstein-data.com/EFTA00485618 p.1. Groff’s reply of 13:49:19 UTC, carrying Nowak’s message in its quoted block, is at https://epstein-data.com/EFTA00485620 p.1 and was read at the image at 600 dpi: the sheet prints “Ah ! Excellent! Thanks”, the space before the exclamation included, and reproduces Nowak’s three lines with the third indented, italic and separated from the second by a double gap, a barred value following it. That layout is consistent with a pasted item. It is reported here as an observation about the sheet and is not the basis of any conclusion on this page, for the reasons given in the body. A second copy of Groff’s reply is at https://epstein-data.com/EFTA02262912 p.1, not read at the image. The italic line reproduces the itinerary line at ⁶². ↩ ↩² ↩³

10. Full-index FTS5 phrase searches of pages_fts.text_content, run 11 September 2026, counting distinct document identifiers, never scoped by documents.dataset. "Caleb Harper" 217 pages / 192 documents (190 EFTA, 2 HOUSE_OVERSIGHT_). scripts/ocr_variants.py "Caleb Harper" --count produced one measured-glyph form, returning 0; --wide produced 32 further single-edit forms, of which "Caleb Harpe" returned 5 pages / 5 documents (one, EFTA02295515, outside the 192) and "Caleb arper" 1 page, already inside. A header-field form reaches 8 pages / 8 documents, three of them outside the 192; those three are not enumerated here. Union of all forms tried: 196 documents. Two further documents reach him under forms no search on his full name finds: EFTA00788080, bylined “C.Harper”, and EFTA00472967, which calls him “Caleb the plant guy”. Forms returning nothing, each run and each returning 0 pages: "Caleb Harpar", "Caleb Harpen", "Caieb Harper", "Cateb Harper", "Ca1eb Harper", "Caleb Harpe r", "CalebHarper", "Harper, Caleb", and the remaining 30 --wide forms. ↩ ↩²

11. Joichi Ito to Jeffrey Epstein, “Re:”, 6 May 2017, Sent: 18:40 UTC, carrying Epstein’s message of 2:23 p.m. local in the quoted block. Read at the image at 400 dpi. https://epstein-data.com/EFTA02388943 p.1. ↩ ↩² ↩³

12. Lesley Groff and Victoria R. Montgomery Brown, “Re: Jeffrey Epstein,” 22 May 2019, 1:24 p.m. to 19:06:28 UTC. Read at the image at 400 dpi. https://epstein-data.com/EFTA00495599 pp.1–2. Second copy: https://epstein-data.com/EFTA02286912 p.1. ↩ ↩² ↩³ ↩⁴

13. Caleb Harper to Jeffrey Epstein, 18 June 2018, 11:32 a.m., with Epstein’s replies of 11:40 a.m. and the exchange of 5 July 2018 above it, all inside the message of 5 July 2018, 15:24:30 UTC. Read at the image at 400 dpi on an undamaged copy, which is the copy quoted. https://epstein-data.com/EFTA00981952 pp.1–2. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

14. Jeffrey Epstein to Caleb Harper, 17 June 2018, 6:10 p.m., quoted in the thread at https://epstein-data.com/EFTA00981952 p.2. Read at the image at 400 dpi on an undamaged copy; the exchange of 5:57 and 6:08 p.m. is on the same sheet. ↩ ↩²

15. Jeffrey Epstein to Caleb Harper, “Re: Scientist,” 3 August 2018, 01:17 UTC, whose quoted block carries the message of 1 August undamaged. Read at the image at 400 dpi. https://epstein-data.com/EFTA02600478 p.1. ↩ ↩² ↩³

16. Caleb Harper to Jeffrey Epstein, “Re: Scientist,” 3 August 2018, 02:19 UTC, carrying Epstein’s “yes,” of 01:17 UTC and the message of 1 August in its quoted block. Read at the image at 400 dpi. https://epstein-data.com/EFTA02600718 p.1. On this copy the damage falls on “th= team” and “=till interested” while “This would be the first guy” and “I’d like you to fund” print clean; on https://epstein-data.com/EFTA02599860 p.1 it falls on “wou=d” and “I’=”, and on https://epstein-data.com/EFTA02600478 p.1 on “first=guy” and “f=nd”. Every word of the sentence is undamaged on at least one of the three, and no single copy is undamaged throughout. ↩ ↩² ↩³

17. Caleb Harper to Jeffrey Epstein, “Project Next Steps,” 24 September 2018, 22:00:17 UTC, with two attachments. Read at the image at 400 dpi on an undamaged copy. https://epstein-data.com/EFTA01022047 p.1. ↩ ↩² ↩³

18. Caleb Harper to Jeffrey Epstein, “Re: Moving forward,” 8 November 2018, 13:33:40 UTC, whose quoted block carries the message of 19 October. Quoted here from https://epstein-data.com/EFTA01017308 p.1, read at the image at 600 dpi, which is the copy that prints the whole message undamaged, sign-off included, and which settles “Ive” against every text layer’s “lye”. Two further copies were read at the image at 400 dpi and both carry the top message damaged: https://epstein-data.com/EFTA02617199 p.1 prints “nuisan=e”, “ex=eriments”, “=roject”; https://epstein-data.com/EFTA02617163 p.1, quoting it inside Epstein’s reply, prints “for=ard”, “hinges =n funding”, “it would =e”. EFTA01017308 is also a fourth undamaged copy of the 19 October message, agreeing with ⁵⁴ on “for support” and “the pipeline with MIT”. ↩ ↩² ↩³ ↩⁴

19. Jeffrey Epstein to Caleb Harper, “Re: Moving forward,” 8 November 2018, 14:05:29 UTC. Read at the image at 400 dpi. https://epstein-data.com/EFTA02617163 p.1. Two further copies carry the header and no body: https://epstein-data.com/EFTA01017180 p.1 and https://epstein-data.com/EFTA01016984 p.1; the body is on https://epstein-data.com/EFTA01017305 p.1. ↩ ↩²

20. Jeffrey Epstein to Caleb Harper, 2 December 2018, 23:49:49 UTC. Read at the image at 400 dpi. https://epstein-data.com/EFTA02611582 p.1. Second copy: https://epstein-data.com/EFTA01014325 p.1. ↩ ↩²

21. Jeffrey Epstein to Caleb Harper, 4 December 2018, 10:40:48 UTC. Read at the image at 400 dpi on two copies: https://epstein-data.com/EFTA01014490 p.1, which prints the line undamaged and is the copy quoted, and https://epstein-data.com/EFTA02611444 p.1, on which “what d[o] you” carries transmission damage. ↩ ↩²

22. All 192 documents in the "Caleb Harper" population were classified from the header block of their own first sheet, with whitespace flattened first and with the sender label matched as From, Fran, Frorn, Frcm, Fram, Front and Fronl, under both of the two header layouts this production uses. Two documents were then corrected by hand after their sheets were opened: one whose parsed sender value had been taken from the wrong field, and one whose sender value is barred but whose body is a message Harper signed. Result: 40 sender, 35 addressed, 112 third-party, 5 without e-mail headers; 40 + 35 + 112 + 5 = 192. The 75 sender-or-addressed documents were read in full. ↩ ↩² ↩³

23. The same sheet, top message: Lesley Groff to Caleb Harper, 28 November 2018, and Harper’s “I’ll be there!” of 27 November 2018, 4:40 p.m. in the quoted block. https://epstein-data.com/EFTA00485579 p.1. ↩ ↩² ↩³

24. Thread “Re: Mtg w/JE”, 11 October 2018, 5:24 p.m. to 6:09 p.m., quoted inside Lesley Groff’s message to Caleb Harper of 28 November 2018, 15:25:24 UTC. Read at the image at 400 dpi; Groff is named in the clear throughout this copy. https://epstein-data.com/EFTA00485579 p.1. ↩ ↩² ↩³

25. "Charles L. Harper" returns 140 documents, "Charles Harper" 100, "Chuck Harper" 47; their union is 198, the figure used on charles-harper. The intersection of that union with the 192 "Caleb Harper" documents is empty, computed as a set difference over resolved identifiers rather than inferred. Bare "Harper" returns 1,099 pages in 834 documents, of which 444 documents lie outside both men; no search on this page was ever run on the bare surname, and no claim here rests on one. Where a sheet used for this page prints only “Harper”, it is a header or byline resolved by its own forename or initial on the same sheet. ↩ ↩²

26. These are Gmail-export sheets. The = is a quoted-printable soft break carried onto the face of the print, replacing one character at a roughly regular interval. The desk has confirmed at the image on other documents that it prints on the page and is not an artefact of the corpus text layer. The damage falls in different places on different copies of the same message, which is what makes cross-copy reading possible: see ¹ against ⁵⁴, and ¹³ against ⁴⁶. Every supplied character on this page is in square brackets and every damaged form is shown as the sheet prints it. ↩ ↩²

27. Caleb Harper, “Re:”, 8 May 2017, 19:01:20 UTC, whose quoted block carries the third party’s message of 2:48 p.m. and Epstein’s of 2:45 p.m. on 6 May 2017. Read at the image at 400 dpi. https://epstein-data.com/EFTA00682099 p.1. ↩ ↩²

28. The three timestamp kinds were checked against each other on matched pairs of copies before any interval on this page was computed. Ito’s reply of 6 May 2017 carries Sent: Saturday, May 6, 2017 6:40 PM on https://epstein-data.com/EFTA02388943 p.1 and Date: Sat, 06 May 2017 18:39:48 +0000 on https://epstein-data.com/EFTA00696018 p.1; Harper’s message of 19 October 2018 carries Sent: Friday, October 19, 2018 6:23 PM on https://epstein-data.com/EFTA02619312 p.1 and Date: Fri, 19 Oct 2018 18:22:52 +0000 on https://epstein-data.com/EFTA01019127 p.1. In both pairs the Sent: field agrees with the UTC header, not with the local time the same thread’s quoted headers print. The 6 May 2017 times used above are: Epstein’s question 18:23 UTC (from the quoted header’s 2:23 p.m. local), Ito’s reply 18:39:48 UTC (RFC header), the introduction 18:45 UTC (quoted header’s 2:45 p.m.), the reply to it 18:48 UTC (quoted header’s 2:48 p.m.). ↩ ↩²

29. The same sheet, top message. https://epstein-data.com/EFTA00682099 p.1. ↩

30. Caleb Harper, “Re:”, 8 May 2017, 19:04 UTC. https://epstein-data.com/EFTA02379094 p.1. ↩

31. Third party to Harper and Epstein, “Re:”, 9 May 2017, 15:36:55 UTC. Read at the image at 400 dpi, which is the basis for the field-state statement in the body: this copy bars the whole sender value and both quoted-header values. https://epstein-data.com/EFTA01046529 p.1. ↩

32. Joichi Ito to Jeffrey Epstein, “Fwd: Mass Spec device,” 7 September 2017, 20:53:11 UTC, forwarding a message of 5 September 2017 from a third researcher to Harper, copied to Edward Boyden and Ito. https://epstein-data.com/EFTA01032604 pp.1–2. ↩

33. Lesley Groff to Jeffrey Epstein, “Joi Ito,” 25 April 2018, forwarding the message above. https://epstein-data.com/EFTA00472869 p.1; https://epstein-data.com/EFTA00472870 p.1. ↩

34. Joi Ito to Lesley Groff, quoted in her reply of 1 May 2018, 10:52:20 a.m., which carries his message of 5:46 a.m. https://epstein-data.com/EFTA02245244 p.1, read from the text layer, not at the image. This document is outside the 192: it names him “Caleb the plan guy” and never “Caleb Harper”, so no search on his name returns it. It was reached by running ocr_variants on the phrase rather than on the name, and it is Ito’s second use of the label in his own voice. ↩

35. Lesley Groff to Jeffrey Epstein, “Reminder Peter Wohlleben the plant guy from NYBG Symposium,” 31 July 2018, 11:43 a.m.; Epstein’s reply of 11:50:37 a.m.; and Groff’s follow-up. Read at the image at 400 dpi on https://epstein-data.com/EFTA02599902 p.1; further copies at https://epstein-data.com/EFTA02252631 p.1 and https://epstein-data.com/EFTA02252419 p.1. "Peter Wohlleben" returns 25 pages in 25 documents corpus-wide; those 25 were not read for this page and nothing is said about them beyond the count. ↩

36. Jeffrey Epstein to Lesley Groff, “Re: Faith Kate’s,” 23 March 2018, 11:36:13 a.m., and a calendar alert of the same date timed 10:00 a.m. Read at the image at 400 dpi on https://epstein-data.com/EFTA02239634 p.1. The alert is at https://epstein-data.com/EFTA02240408 p.1 and was not read at the image; it is quoted from the text layer. ↩ ↩²

37. Lesley Groff to Martin Nowak, “Re:”, 27 November 2018, 13:45:11 UTC, carrying Nowak’s message of 8:42 a.m. Read at the image at 400 dpi; the marks around “plant guy” are the sheet’s own. https://epstein-data.com/EFTA00485614 p.1. Second copy: https://epstein-data.com/EFTA02262786 p.1. ↩ ↩²

38. Delivery-failure notice forwarded to John De La Parra, 29 November 2018, 3:58:02 PM, carrying the invitation of 27 November 2018, 9:10:00 AM EST at its foot. Read at the image at 400 dpi, both sheets. https://epstein-data.com/EFTA02265833 pp.1–2. The sender is barred in the header of this copy and the invitation’s signature block prints the role without the name; the attribution to Lesley Groff rests on her own message of 28 November at ²³, in which she reports having received the bounce. ↩ ↩² ↩³

39. Two iMessages, both Sender: Self from Jeffrey Epstein’s own handset, 20 October 2018, timestamped 17:28:01 and 17:28:42 and labelled (UTC) on the sheet itself. Read at the image at 400 dpi; the sheet prints each message twice, in duplicate log blocks, and the participants other than the sender are barred. https://epstein-data.com/EFTA00782011 p.14. A second copy of the same exchange, carrying an intervening reply, is at https://epstein-data.com/EFTA00508611 p.16, read from the text layer. The name is spelled “John de la para”, dropping the second r: "John De La Para" returns this one page and nothing else, so no search on the correct spelling reaches it. This page did not hold this document until a third gate found it, and it is the most direct evidence on the page of what Epstein thought of the man the money was for. ↩ ↩²

40. Ten search forms were run, not one. "Appt w/Caleb Harper" returns 11 pages; "Appt w/Caleb" returns 13, the two extra carrying the surname OCR-truncated as “Appt w/Caleb Ha” (EFTA02262537) and “Appt w/Caleb Harpe” (EFTA02262756, which this page already cites at ⁷). "Appt w Caleb" and "Appt w/ Caleb" return the same 13; "2:00pm Appt w/Caleb" returns 12; "Apptw/Caleb" and "Appt wiCaleb" return 0. Union: 13 pages in 13 documents, all opened. Every one is a copy of the itinerary for Sunday 2 December 2018. The copies are not identical and the day was still moving: of the thirteen, nine carry a George Church appointment and four do not, eight carry a Seth Lloyd lunch and five do not, and two carry neither Nowak nor Church. What does not vary is the pair: on twelve of the thirteen the De La Parra line stands immediately above the Harper line, and on the thirteenth the match falls on a second sheet whose preceding lines are on the page before. Each was day-resolved from an anchor on its own document, not from block identity: eleven carry a printed “Sun. Dec. 2, 2018” heading above the block; EFTA02265760 dates itself (“Sent Thur 11/29/2018”, asking about “anyone else you should see this Sunday”); and EFTA02264659, whose match is on p.2 with no heading above it, is settled by its own p.1, which carries “Sent: Fri 11/30/2018”, the subject “Re: Harvard Trip Itinerary”, and the “Sun. Dec. 2, 2018” heading over a block that straddles the page break. Thirteen of thirteen. None was read at the image. He is also listed for 6 May 2018, but inside a block of time with Ito (“Confirmed Caleb Harper (the plant guy)”) rather than as an appointment in his own name, so the claim here is about appointments in his own name and not about whether the two men had met before. ↩ ↩²

41. Joi Ito to Jeffrey Epstein, 19 February 2019, 15:23:24 UTC, carrying Epstein’s question of 15 February in the quoted block. Read at the image at 400 dpi. https://epstein-data.com/EFTA01030310 p.1. This document is cited on joi-ito and is not a document about Caleb Harper; it is referred to here only to mark an open question and to keep it where it belongs. ↩ ↩²

42. Caleb Harper to Jeffrey Epstein, “Book on sentient plants,” 6 May 2018, Sent: 17:44 UTC. Read at the image at 400 dpi. https://epstein-data.com/EFTA02670367 p.1. ↩ ↩²

43. The header of https://epstein-data.com/EFTA00981952 p.1 prints Inline-Images: image_67201793.JPG; read at the image at 400 dpi. On https://epstein-data.com/EFTA02575858 p.2 the quoted header for the message of 17 June 2018, 2:31 p.m. is followed by an HTML image tag carrying a cid: reference and a width attribute, and by no text; that sheet is read from the text layer, not at the image. No copy of this thread read for this page carries a body for that message. ↩

44. The same thread at https://epstein-data.com/EFTA02580515 p.2, where the extraction carries the quoted header of the 6:10 p.m. message but not its body. The message survives on the copy at ¹⁴, which is why it is quoted from there. ↩

45. FTS5 phrase searches over text_content, 11 September 2026, intersected with the 192-document population by resolved identifier. "500k" 1,243 pages / 1,144 documents corpus-wide, 24 inside the 192, all of them resolved and every one a copy of the 18 June budget thread or the October-November “Moving forward” thread. "500,000" 8,197 documents corpus-wide, 0 inside; "1.5 million" 499, 0 inside; "1,500,000" 1,862, 0 inside. The corpus-wide counts are given to show the searches are not empty; only the intersections bear on the claim. ↩

46. The same passage carrying transmission damage at https://epstein-data.com/EFTA02580515 pp.1–2 and, re-sent in September under the heading “Previous[ ]email breaking down funding”, at https://epstein-data.com/EFTA02619217 pp.1–2 (read at the image at 400 dpi). Only https://epstein-data.com/EFTA02619217 prints the physiologist line’s overhead figure as “+5=%”; https://epstein-data.com/EFTA02580515 prints “+56%” clean at that point and is damaged elsewhere on the same line. EFTA02619217 also prints ”= total cost” where the undamaged copy prints “k total cost”. ↩ ↩²

47. Caleb Harper to Jeffrey Epstein, 6 September 2018, 7:57 p.m., quoted in his message of 12 October 2018, 12:46 a.m. Read at the image at 400 dpi. https://epstein-data.com/EFTA02619217 p.1. Further copies at https://epstein-data.com/EFTA01020029 p.1, https://epstein-data.com/EFTA01023259 p.1 and https://epstein-data.com/EFTA01024180 p.1. ↩ ↩² ↩³

48. 80,000 x 1.20 = 96,000; 96,000 x 1.56 = 149,760, which is the figure the line itself prints. 100,000 x 1.20 x 1.56 = 187,200 (“187k”). 40,000 x 1.20 = 48,000. 100,000 x 1.20 = 120,000. Sum: 504,960. The arithmetic is the document’s own and is reproduced here as a check on the reading of the damaged digit, not as a finding. ↩

49. Caleb Harper to Jeffrey Epstein, “Re: Scientist,” 1 August 2018, 17:04 UTC Read at the image at 400 dpi. https://epstein-data.com/EFTA02599860 p.1. ↩ ↩²

50. Merlin Sheldrake to Jeffrey Epstein, “papers,” 17 July 2018, 7:07 a.m., forwarded by Epstein to Harper at 7:24 a.m. and surviving in the quoted block of the “Scientist” thread. https://epstein-data.com/EFTA02600478 p.1; https://epstein-data.com/EFTA01005608 p.1. ↩

51. Jeffrey Epstein to Caleb Harper, 31 August 2018, 17:03:42 UTC. Read at the image at 400 dpi. https://epstein-data.com/EFTA02625953 p.1. ↩

52. Jeffrey Epstein to Caleb Harper, “Re: Project Next Steps,” 25 September 2018, 20:12 UTC, and Harper’s reply at 20:16 UTC. https://epstein-data.com/EFTA02621689 p.1; https://epstein-data.com/EFTA02621671 p.1. ↩

53. “Plant Sentience”, C.Harper and J. de la Parra, dated 9.22.18, one sheet. Read at the image at 400 dpi; the bold emphasis quoted is the document’s own. https://epstein-data.com/EFTA00788080 p.1. "plant dictionary" returns this document and no other; "CHarper" likewise. ↩

54. A second copy of the same message, in a different rendering, carrying no damage at either of the two points quoted. Read at the image at 400 dpi. https://epstein-data.com/EFTA01019127 p.1. ↩ ↩² ↩³

55. The 2014 sum is described at joi-ito, sourced there to Ito’s message to Epstein of 29 May 2014 (“I just got notice that $500K came into my slush fund account”) at https://epstein-data.com/EFTA00990778 p.1. It is a different sum, in a different year, on a different account, and no document read for this page connects the two. The comparison is drawn here only to prevent the two being merged. ↩

56. FTS5 conjunctions over text_content, run 11 September 2026. Returning zero pages: "Caleb Harper" AND "wire", "Caleb Harper" AND "500", "Harper" AND "wire transfer", "gift information" AND "Harper", "sentience" AND "gift", "De La Parra" AND "500". "Caleb" AND "wire" returns 3 pages in 3 documents; "Caleb" AND "payment" 10 pages in 10 documents; the sets share one HOUSE_OVERSIGHT_ identifier, so the union is 13 pages in 12 documents, and all 13 pages were opened and read. They are: a 2001 letter addressed to a court in the I. Caleb Boggs Federal Building; a 2025 news item naming an immigration official; a court-locations notice naming the Joseph Caleb Center, which sits in a corpus row holding an entire 99-sheet run and is therefore not page-pinnable; three 2010 construction-payment messages about a contractor named Uriel Caleb; two bank statements in which “Caleb” is scanner noise; a securities schedule in which it is scanner noise; two copies of a 2018 Bloomberg article by a reporter named Caleb Melby; and a travel disclosure carrying the string “CALEB:EWA”. None concerns this Caleb. ↩

57. "landed at ML" returns 4 pages in 4 documents — EFTA01020029, EFTA01023259, EFTA01024180 and EFTA02619217 — all four of them copies of Harper’s own message of 6 September 2018 asking whether it had happened, and all four were opened. "has landed" returns 191 pages in 189 documents; that set was not read and is reported here as a bare count with no characterisation of its contents. "Caleb Harper" AND "funding" returns 29 pages in 29 documents and "Caleb Harper" AND "fund" 17 in 17; both sets lie inside the 192 and were opened as part of the 75 read in full. ↩ ↩²

58. The office day sheet is https://epstein-data.com/EFTA00285585 p.5, which heads the day Sat Oct 20, 2018 PB. The sheet carries a location marker on each dated heading and the marker varies across the six pages, PB 18 times, NY 4 and LSJ once, so it is tracking where he was rather than where the office was; the preceding day’s block prints 6:37pm Depart NY and 8:47pm Arrive PB. Against that, the handset messages at ³⁹ are labelled (UTC) on their own sheet at 17:28, which is early afternoon Eastern on the same Saturday. Both are records of his own office’s making and they do not agree. Nothing here resolves which is right, and neither bears on Harper, who is absent from both. Read from the text layer. ↩

59. Jeffrey Epstein to Joi Ito and Martin Nowak, 6 October 2018, 6:05 p.m., quoted in the thread. https://epstein-data.com/EFTA01020583 p.1. ↩

60. The confirmation of 27 November 2018, 11:29:53 PM. Read at the image at 400 dpi; this copy prints the street number as a digit where other extractions render it as a capital I. The sender’s name is barred on this copy. https://epstein-data.com/EFTA02262634 p.1. Further copies: https://epstein-data.com/EFTA00485023 p.1; https://epstein-data.com/EFTA00485939 p.1. ↩

61. John de la Parra to Lesley Groff, copied to Harper, “Re: Jeffrey Epstein,” 2 December 2018, 1:04:56 PM, carrying Groff’s reconfirmation of 1 December. https://epstein-data.com/EFTA02265838 p.1. ↩

62. Itinerary for 15 October 2018: “2:00pm Appt w/John De La Parra (from Caleb Harper’s group)”. The phrase "Caleb Harper's group" returns 33 pages, and the damaged form "Caleb Harper's roup" adds one more: union 34, all opened — 32 copies of the 15 October itinerary line and 2 Nowak’s reply of 27 November, which reproduces the naming portion of that line without its time prefix. Read at the image at 400 dpi on https://epstein-data.com/EFTA02261619 p.1; further copies at https://epstein-data.com/EFTA02259929 p.1, https://epstein-data.com/EFTA02260037 p.1, https://epstein-data.com/EFTA02260490 p.1, https://epstein-data.com/EFTA02261228 p.1 and https://epstein-data.com/EFTA02261680 p.1. ↩ ↩²

63. "John De La Parra" returns 85 pages in 85 documents; the surname alone, "De La Parra", returns 90 pages in 88 documents, and the two figures are the full-name and surname forms of the same search, not two populations. Six were opened and read in full for this page — EFTA00485023, EFTA00485579, EFTA00485939, EFTA02261619, EFTA02262634 and EFTA02265838. The remaining 79 were counted and not read, and nothing is said here about their contents. Everything this page says about De La Parra comes from the six, from the “Scientist” thread, and from the messages of 6 September and 2 and 4 December quoted above. ↩

64. Every date token on every sheet of all 192 documents was extracted and sorted, in four written forms. No document is undated. Distribution of each document’s latest date: 2017 11 documents, 2018 178, 2019 3. The three 2019 documents are EFTA00495599, EFTA02286597 and EFTA02286912, all copies of the exchange of 22 May 2019; the two documents dated later than 4 December 2018 and earlier than that exchange are both day sheets from Epstein’s office, not correspondence. A first run of this measurement scanned only the sheets carrying his name and reported eight undated documents; that was a property of the scan and not of the release, and it is recorded here because the corrected figure is the one the claim rests on. This bounds the claim to the 192 documents, and to nothing wider. ↩

65. "OpenAg" returns 23 pages in 19 documents; "Open Agriculture" 15 pages in 13 documents; union 21 documents, all opened. Twenty carry the phrase in Harper’s own signature block, in a link to the initiative’s community forum, or in a scheduling line (“Caleb (OpenAg)”, “openAg head”, “plant pheromone scientist, openAg head, Ed Boyden”). The twenty-first is EFTA00788080, the “Plant Sentience” paper. "food computer" and "personal food computer" each return 0 pages. ↩