Every article here that makes a dated claim about the future gets an entry on a public scorecard. Each entry carries three things: the date it was made, a test sharp enough to settle it, and a status. The reason for the page is written at the top of it. A record is only worth keeping if being wrong costs something.
The first entry ever to come due came due this month. It was filed on 15 May 2026 and read like this:
The Q2 2026 13F (due ~14 August 2026) shows IBIT and FBTC holdings at or below the Q1 levels. A material increase resolves this a miss.
IBIT is BlackRock's bitcoin fund. FBTC is Fidelity's. A 13F is a form big investors must file with the SEC every three months, listing what they held on the last day of the quarter. Jane Street filed its June form on 14 August. So the call is due, and this article is the grading of it.
It turned into something else on the way. This article does four things. It shows that the reason the May call gave for itself is false, and that the filing said so on its own cover page. It splits the dollar change in these holdings into the part somebody chose and the part that was only the new price. It grades the entry, and the grade is a miss. And then it turns that same method on itself, because one row of the March form was out of reach for six days, and reading it reversed the sign of the section above.
The Sentence That Was Wrong
The May article gave a reason for its own doubt, and the reason was this: a 13F shows only the long leg. Put plainly, it said the form lists the shares a firm owns and nothing else. A falling share count might mean a trade was being closed out. It might equally mean one of four other things, and the article listed all four.
That reason is false, and both filings say so in their own text. Every row of the form is tagged as shares, as a call, or as a put. Calls and puts are the two kinds of option contract: the right to buy something, and the right to sell it. Jane Street's June filing has four rows against BlackRock's fund: shares, calls and puts. Its March filing has seven rows against the same fund, four of them options, and seven against Fidelity's, four of those options too. Grayscale's two trusts carry five rows each in March, options included.
The March rows against BlackRock's fund are the ones that matter here, because BlackRock's fund is what the May sentence was written about. They were the rows out of reach, so for six days the point had to lean on the three funds beside them. It no longer does. In March the firm reported 45,287,400 shares' worth of calls against that fund and 44,212,100 shares' worth of puts, beside 5,872,212 shares held outright. Each option line is more than seven times the size of the holding of shares next to it.
The firm said so on the cover page too. The same sentence sits on both filings, word for word: the report "includes options that deliver two different securities and each has been reflected separately". Nothing was hidden. It was the part of the filing nobody read.
Two things follow, and they are not the same thing. The first is that the reason given in May does not hold. The second is not that the call was wrong. The December filing has never been opened here, and December is the quarter the May piece measured its fall from. So the argument is left standing without the document it rested on, which is not the same as being shown false. What May described may well have happened. This document does not show it. That gap is the whole of the correction.
What the Option Rows Do Not Say
Three limits, each of which a careless sentence would break.
A 13F reports what is held. Options a firm has sold, and shares it has borrowed and sold short, never appear on it at all. So the call rows and the put rows are both things the firm owns. There is no short side in this document and there never was one.
The number on an option row is not exposure. In June the firm reports 34,033,900 shares' worth of calls and 34,804,000 shares' worth of puts against BlackRock's fund, beside 24,878,191 shares held outright. Those figures are not a hedge. They are not an offset, and they are not a flat book. Nothing in the form lets a reader work out what they are. No net position is calculated here because none can be.
And nobody has yet checked what the SEC's own rule book asks a filer to report about options. Until someone does, the honest move is to say what the rows say, not what the rules want.
Four Funds, Four Different Directions
Set the option rows aside and look only at shares held outright, which is what the call was about.
Fidelity's fund went from 1,954,174 shares in March to 886,736 in June, a cut of 54.62%. Grayscale's main trust went the other way, from 263,972 to 340,363, up 28.94%. Grayscale's smaller trust held 194,603 shares in March and has no share row at all in June. That holding is gone whole, though its option rows remain. BlackRock's fund went the same way as Grayscale's main trust but far harder: 5,872,212 shares in March, 24,878,191 in June, a rise of 323.7%.
Two more funds turn up in June with nothing to match them against: ARK's at 2,700,151 shares and Bitwise's at 421,020. Invesco's bitcoin fund is missing from the June form, and that gap was measured rather than assumed. The rows run from Invesco's exchange traded fund trust straight to its ethereum fund with nothing in between. A bitcoin row would have to sit between the two, by name order. There is no space there. The March form was never searched for it, so this is an absence and not an exit.
Set all of that against the book as a whole. The firm reported $777,217,056,890 of holdings in March and $1,205,633,745,715 in June, a rise of 55.12%. The row count went the other way: 19,058, then 14,718. A bigger book is not lifting these lines, and it is not sinking them either.
So four funds moved four different ways in the same quarter. One halved. One grew by close to a third. One was dropped whole. And the largest of them, the one this whole grading turns on, went up more than fourfold. A firm acting on a single view about bitcoin does not do that. What it was doing instead is written nowhere in the document, and this article does not guess.
A Quarter of the Fall Was Nobody's Decision
Write-ups of a 13F report these holdings in dollars. Dollars move for two reasons that have nothing to do with each other. The shares get re-priced at the new quarter's closing price. And the number of shares changes. Only the second is a decision by anyone.
The form makes the split easy. Every row carries a share count and a dollar value, so dividing one by the other gives the price the filer used. That price comes out the same on every row of a given fund, which doubles as a check that the rows were read right. Fidelity's fund prices at $59.03 in March across all seven of its rows, and $51.05 in June across all four.
All four funds were re-priced down by nearly the same amount: 13.35% for BlackRock's, 13.52% for Fidelity's, 13.72% for Grayscale's main trust, 13.47% for the smaller one. Four pairs, copied out one at a time, landing within four tenths of a point of each other. BlackRock's was read by a different tool from the other three and it lands in the middle of them, which is the best evidence available that all four were read right.
Over the same two dates bitcoin itself fell less. This site's daily file has it closing at $66,699.27 on 31 March and $60,152.36 on 30 June, down 9.82%. The four fund prices fell 3.70 points further than that on average. No cause is offered for the gap here. A daily close and a fund's quarter-end share price are struck at different hours, on different calendars. The gap is reported, not explained.
Now split the dollars. Start with the three funds whose March and June rows could both be read when this section was first written. Taken together they fell from $135,118,198 to $60,761,197. Of that $74.4m drop, $18.3m is the re-pricing. The other $56.1m is shares truly leaving. So roughly a quarter of the fall is a number nobody at the firm decided. Had the share counts not moved at all, those lines would still have printed 13.54% smaller.
Grayscale's main trust is the sharpest case, because there the two pull opposite ways. Its dollar line rose 11.25%. Its share count rose 28.94%. A reader who follows the dollars sees less than half of what was decided.
None of that changes the direction of any of those three lines. It changes the size, and in one case it hides most of it.
Then the Fourth Row Reversed It
The fourth row was read on 22 August, six days after the section above was written. It was never hidden and it was never missing. The March list of holdings runs to 11,162,326 bytes, the tool sent to fetch it stopped after five megabytes, and BlackRock's row sits at about the halfway mark. A second tool built to stream the whole document read it end to end and matched the row by its security number.
Put that row back in and the paragraph above does not survive it. The three matched holdings fell 55.03%, from $135,118,198 to $60,761,197. All four rose 146.43%, from $360,728,583 to $888,956,176.
The method did not change and neither did the three figures. They are still true of those three lines. What changed is what they were evidence about. Split the four the same way and $48.4m of the move is re-pricing against $576.6m of shares. BlackRock's line alone carries $632.7m of share buying against $30.1m of re-pricing, which is more than the whole of the rest put together, in both directions at once.
So three of the four rows said this book more than halved. All four say it grew to nearly two and a half times. This article was written to show what reading part of a filing costs. It then paid that cost itself, in its own middle section, on the one row it had already named as unread.
One more reading the four rows rule out. Between March and June the option lines against BlackRock's fund got smaller, not larger. Calls fell 24.8%, from 45,287,400 to 34,033,900. Puts fell 21.3%, from 44,212,100 to 34,804,000. Over the same quarter the shares held outright rose 323.7%. Whatever was going on, the options book did not grow alongside the position. That is a statement about two reported numbers and nothing more. It is not a hedge, not a measure of exposure, and not a motive.
The Grade Is a Miss
The entry named two funds. It needed both to sit at or below their March levels.
The Fidelity half did what the call said it would. Down 54.62%, read from the filings on both sides. That half is a hit.
The BlackRock half is the one that settles the entry. In March the firm held 5,872,212 shares of that fund, worth $225,610,385. In June it held 24,878,191, worth $828,194,979. That is 19,005,979 shares more than it started with.
This is not a case of one half up and one half down, therefore nothing to say. The test names two funds and needs both. So a single failing half settles it as a miss, with no sums required. A holding four times its old size is a material increase on any reading, and a material increase is the miss trigger the entry wrote for itself. So the entry is a miss. Nothing about that needed a judgment call or a threshold to argue over. Why the holding grew is not in the document, and no reason for it is offered here.
It is worth saying plainly what the six-day wait was and was not for. On 16 August this piece declined to grade the entry against news write-ups of the March filing, which put the holding near 5.9 million shares. The filing says 5,872,212, so those write-ups were close and the direction would have been the same. The wait was never a doubt about the figure. It was that an article about a call graded on a document read in part cannot settle its own entry the same way and keep a straight face. The grade now rests on the filing, which is the only thing it could rest on.
The May article is corrected in place, and the note says the reason it gave was wrong. It does not say the conclusion was wrong, because that has not been shown. The December filing is still unopened, and December is the quarter that argument measured its fall from. What is being corrected is not a bad guess about the market. It is the part of the filing nobody read.
Three Notes for Anyone Repeating This
Sites that republish 13F data have Jane Street holding 11,403 positions in March and 11,385 in June. The filings' own cover pages say 19,058 and 14,718. The dollar totals agree to the dollar, so nothing was mis-read. A "holdings" count on a data site is a choice about how to group rows, not a count of the document. One such site also folded the June BlackRock holding into a single line. The filing splits it across two arms of the firm.
Every row of a given fund works back to one price. Fidelity's March rows do it seven ways, BlackRock's March rows seven ways, Grayscale's two March trusts five ways each, BlackRock's June rows four ways. That check is what makes the figures above worth printing rather than merely quoting.
And the fund whose row could not be reached was reachable in every other way. The document answers every request in full. The limit was a tool, not a wall, and the difference matters. A row nobody could read is not a row that is not there. It has since been read, and all seven of its March lines price at $38.42.
What It Cost
Four things came out of this that were not known before it. A 13F does not list only the shares a firm owns, and the option rows sit in the March filing the May piece was built on. Roughly a quarter of the dollar fall in the three matched holdings was a re-pricing nobody chose, and in Grayscale's main trust the dollars hide more than half of what was decided. The first entry on this scorecard is a miss, graded on the filings and on nothing else. And the row that took six days to reach turned the middle of this article upside down.
A scorecard exists so that a claim costs something. The first entry to come due has cost more than a grade. It has cost a sentence in the article that made it, and that sentence is now struck. It has also cost this article a finding of its own, which is a fair price for publishing one about other people's partial reads.
The number that finished the grading was one row of a public document that anyone may download. It was never hidden. It was the part of the filing nobody read.
Where the figures come from: the holdings are read from the two filings themselves, SEC accession numbers 0001595888-26-000043 for the first quarter and 0001595888-26-000108 for the second. Every price is a daily close from this site's own price file, which ends on 21 August 2026. One figure in this article came from news write-ups rather than the filing, and it appears only as the thing the filing replaced.
Tracked: the Jane Street entry this article grades is logged on the predictions scorecard, where it is now recorded as a miss, with the dated resolution note explained above.