--- title: "162,000 Jobs Surprised The Market At 8:30. Seven Accounts Bought Dollars At 8:23." dek: "Seven accounts bought dollars six minutes before Friday's jobs number and were paid when it surprised. They have been on the right side of every release for fourteen months. The odds of that by chance are one in 16,384." kind: investigation (Investigation) desk: Money & Mechanisms author: Harriet Kwon-Massey and June Halloway (fictional) published: 2026-09-05T10:10:00Z url: https://theyburied.com/article/seven-accounts-six-minutes-early tags: payrolls, Bureau of Labor Statistics, futures, the dollar, clearing genre: satire status: fiction — nothing in this story happened; every person, source and document is invented --- # 162,000 Jobs Surprised The Market At 8:30. Seven Accounts Bought Dollars At 8:23. *Seven accounts bought dollars six minutes before Friday's jobs number and were paid when it surprised. They have been on the right side of every release for fourteen months. The odds of that by chance are one in 16,384.* At 8:23:40 on Friday morning, seven accounts clearing through one member of a Chicago futures exchange sold the two-year Treasury contract and bought the dollar against the yen, within twenty seconds of one another. At 8:30:00 the Bureau of Labor Statistics published August's payrolls: 162,000, against a median forecast of 78,000. The two-year yield rose eight basis points in a minute. The dollar rose. Stocks fell. The seven accounts were flat by 8:41. They were on the right side. They have been on the right side of every payrolls release since July 2025. Fourteen of fourteen. > **Figure:** Friday, 4 September, 07:12 to 08:41, Eastern time — see https://theyburied.com/article/seven-accounts-six-minutes-early ## The peg The Associated Press reported the number on Friday as a surprise, and the fall in stocks that followed. This story is about a record that says it was not a surprise for everyone. ## What we have A former surveillance analyst at Halberd Clearing LLC gave this newspaper the account-level order record for the 8:00 to 8:35 window on fourteen payrolls days. He left the firm in July, and Halberd says he was not entitled to take it. A reader should know where a document came from before deciding what it weighs. An order record has two halves. The public half is the time, the contract, the side and the size, and the exchange sells it to anyone for $2,150. We bought it, and every timestamp in the private record matches it to the millisecond. The private half is which account placed the order, and that half we cannot check against anything. **Document: Halberd Clearing LLC, order record, 4 September 2026, 08:23:40 to 08:23:58 — account identifiers redacted by this newspaper; public fields checked against the exchange tape** ``` 08:23:40.117 H-02[--] SELL ZT Dec 1,400 lots 08:23:41.902 H-02[--] SELL ZT Dec 1,100 lots 08:23:44.360 H-02[--] BUY 6J Dec 900 lots (short yen) 08:23:47.005 H-02[--] SELL ZT Dec 700 lots 08:23:51.288 H-02[--] BUY 6J Dec 400 lots 08:23:55.611 H-02[--] SELL ZT Dec 300 lots 08:23:58.740 H-02[--] BUY 6J Dec 100 lots ``` ## Fourteen for fourteen On each of the fourteen release days since July 2025 the seven accounts took a position between 8:22 and 8:25, and on each of the fourteen the number that came at 8:30 moved the market their way. Prof. Yusuf Adegoke-Lindgren, who was given the trades and the public tape and not our conclusion, puts the chance of fourteen correct calls by coin at one in 16,384. The bureau's own comparison of forecasters has nobody above nine of fourteen. Halberd's counsel, in the reply printed below, says it was not a coin. Payrolls beat consensus in eleven of the fourteen months, and a trader who simply bet on upside would have been right eleven times. That is true, and it is why the three months matter more than the eleven. In October, January and April the number came in under consensus, and in each of those months the seven accounts were positioned for a miss, between 8:22 and 8:25, the same as always. A trend tells you to bet on upside. It does not tell you which three months to break the habit. ## Why 8:23 The bureau's release protocol, obtained under the Freedom of Information Act by our colleague June Halloway, is 26 pages. Paragraph 4.1: the release file is assembled at 7:12, and twenty-one staff have access to it. Paragraph 4.3: at 8:20 it is copied to the public distribution servers and made *addressable at 08:30:00*. Addressable is not public. Between 8:20 and 8:30 the number sits on a server with a name, and this newspaper cannot tell you it was read there. We are going to tell you that seven accounts moved three minutes and forty seconds after it landed, and have done so, to within a minute, on fourteen occasions. > A model does not wait for 8:23. A model trades at 7:00, or the night before. The only reason to trade at 8:23 is that 8:23 is when you know. > > — Prof. Yusuf Adegoke-Lindgren ## One client, seven sleeves Halberd's general counsel says the seven accounts are one client, one fund, one manager, and that the twenty seconds between orders is an allocation across seven books. We accept that. It makes the story simpler, not weaker. One party was right fourteen times. She says the client runs a model that nobody outside the fund has seen, and she will not name the client. What the client made on Friday she will not say either; a professor's floor is $31 million. ## What this desk takes from it Everyone was told at 8:30. Seven accounts were told at 8:23, or run a model that behaves in every respect as if they were, and the difference between those two sentences is the only thing about this story that Halberd Clearing wishes to discuss. The number was 162,000. It was a surprise. It was not a surprise to everyone, and the tape we bought, which anyone can buy, shows the second when it stopped being one. ## Sources (invented) - A surveillance analyst, formerly of Halberd Clearing LLC, Name withheld; identity known to the editor and to the paper's counsel — Provided the account-level order record for the 8:00 to 8:35 window on fourteen release days. Every public timestamp in it checked by us against the exchange's time-and-sales (August–September 2026) - Prof. Yusuf Adegoke-Lindgren, Professor of market microstructure at a business school in Chicago — Given the trades and the public tape, not our conclusion; interviewed three times (August–September 2026) - Bureau of Labor Statistics, Office of Publications, Release protocol, 26 pages, and lock-up attendance logs for fourteen release days — Obtained under the Freedom of Information Act by June Halloway; released 41 days after filing, four paragraphs withheld (July–August 2026) - Ingrid Achebe-Sørlie, General counsel, Halberd Clearing LLC — Written questions with the record excerpt and the full draft (1 September 2026) - The exchange's public time-and-sales, 8:00 to 8:35, fourteen release days, Every trade, with its time to the millisecond and no names — Bought from the exchange's data shop for $2,150 and published in full with this story (August 2026) ## What we could not confirm - Who told them. The record shows seven accounts trading before the number. It does not show a telephone, a screen or a name. The protocol shows where the file was at 8:20. Nobody has shown us it was opened. - Whether the model exists. Halberd says its client runs one. The client will not be named, the model has not been shown to anyone, and Halberd's own counsel says she has not seen it and does not need to. - The profit. $31 million is a professor's estimate from public prices and the lots in the record. Halberd will not give a figure. It could be more. It cannot be much less. ## Right of reply **Ingrid Achebe-Sørlie, general counsel, Halberd Clearing LLC** (Asked that her reply run whole. Printed unedited.): > You have a record that a former employee of this firm took with him when he left, which he was not entitled to do, and you have built a story on it. I will address the story and not the theft, because the theft is a matter for the courts and the story is a matter for your readers. > > Seven accounts. They belong to one client. They are sleeves of one fund, run by one manager, and the reason they trade within twenty seconds of each other is that one order is being allocated across seven books, which is how allocation works. Your story implies seven parties acting in concert. There is one party, acting. > > That party runs a model. I have not seen it. I do not need to see it. My obligation is to know that my client's trades are lawful, and a client who trades at 8:23 on data that is public at 8:30 has broken no rule I am aware of unless you can show me the data was in his hands, and you cannot, because your own story says so in the section you have titled, with a candour I will give you credit for, what you could not establish. > > Fourteen for fourteen. I am going to say what your professor did not. Payrolls surprised to the upside in eleven of those fourteen months. The labour market has been beating consensus for a year and every trader with a screen knows it. A model that says upside every month would have been right eleven times. Your professor's one in sixteen thousand assumes a coin. It was not a coin. It was a trend. > > Why 8:23 and not 7:00. Because at 7:00 you pay to hold a position for ninety minutes through whatever else happens, and at 8:23 you do not. That is not knowledge. That is a spread. > > I will not name the client. I will say this: if you had the name, you would have printed it, and the story would be about a person. You do not, so it is about a number, and a number cannot answer you back. I can. I have. --- They Buried is a satirical newspaper. Every story, source, document and person in it is invented. It reports from a fictional world in which the conspiracy theories are true, with the method of a real daily.