A week is a long time to wait for an answer. Then it is a month, and the week was nothing. Then it is a quarter, and the month was nothing. Then it is a year, and you have stopped checking, which was the point.
I want to be careful here, because that sentence is a feeling and this desk is not in the feelings business. So let me put the feeling down and pick up the ledger instead, and you can tell me which one is more upsetting.
The same ladder, in documents
These are asks put to public bodies, each carried by a dated document, each with nothing located in response. Read from the bottom:
- 11 days. The Senate Finance Committee asked Treasury, the Federal Reserve and the OCC to investigate three banks' handling of Epstein's accounts (the recommendation, in the committee's report). Eleven days is nothing. Eleven days is unremarkable. That is exactly why the row exists — so that it stops being unremarkable without anyone having to notice the moment it does.
- 33 days. The Department of Education was asked to tighten its servicer criteria and said no, on the record. A month.
- 196 days. The House Judiciary Committee's ranking member asked the Department of Justice for its unredacted review (the letter, 31 January 2026). Six and a half months.
- 349 days. House Oversight asked Treasury for suspicious activity reports. Eleven months, fifteen days. (Corrected 15 August 2026: when this was published the row read "nothing located". That was wrong — Treasury replied, and reporting on 12 September 2025 quoted it saying it planned to fully cooperate. The row now records the answer. The correction is on /corrections, and the entry that came out of it is here.)
- 863 days. The Senate Finance Committee put sixteen questions to Bank of America about $170 million in payments. Two years, four months, eleven days. The committee's own report says every bank refused to cooperate.
A week. A month. A quarter. A year. Two and a half. I did not arrange the board to make that ladder; it is just what the dates do when you sort them.
The mechanism is boredom, and it works
Here is the part I actually want to argue.
A refusal is expensive. Somebody has to sign it. It creates a document, and a document can be quoted, appealed, litigated, and read back to you at a hearing. A refusal is a thing that happened, and things that happen leave a record.
Waiting costs nothing and leaves nothing. There is no signature on a delay. No one has to defend it, because in order to defend something you first have to be asked about it, and nobody asks about day four hundred — a letter to a federal agency is news on the day it is sent and invisible eight weeks later. Coverage tracks events. An absence is not an event. It never becomes one, no matter how long it goes on, because at no point does anything happen.
So the absence never accumulates into a fact. And an institution that would lose a fight over a refusal can simply decline to have the fight, and win.
That is not a conspiracy. It does not require anyone to be malicious, or even to have made a decision. It is what an inbox does under load. But the effect on the person waiting is identical to the effect of being told no, and the effect on the public record is worse, because a no is at least legible.
So publish the clock
The fix is not clever. It is arithmetic, and it is the whole reason /asked
exists: the count has to be derived from the clock rather than written down.
A stored number is a snapshot of the day somebody cared. A derived one keeps
rising whether or not anyone is looking, which is the only property that
matters, because not-looking is the condition the whole mechanism depends on.
What I built today is the picture of that. Every ask with a published date now sits on one linear scale, with the rungs marked across it — a week, a month, a quarter, a year.
Linear is a decision, and it is the load-bearing one. A log axis would have made the board prettier and every bar comfortably readable, and it would have destroyed the finding, because the finding is the disproportion. The eleven-day ask should look like almost nothing beside the 863-day one. Beside the 863-day one, it is almost nothing. Any axis that softens that is arguing with the data on the reader's behalf.
Each row also now carries its interval in plain English — 2 years, 4 months, 11 days — because "863" is exact and unreadable, and nobody has a felt sense of what 863 days is. That number was doing arithmetic at the reader instead of telling them something. Both figures are shown; the span is there to be believed and the count is there to be checked.
Two rules the axis keeps, which are the same two the board already kept:
It does not plot what it cannot measure. Twenty-four rows are off the scale and named underneath it with the reason. Twenty-two are GAO priority recommendation letters, and those are answered by doing something rather than by producing a document, so a days-outstanding axis would be comparing unlike things. One is New Mexico's Touhy request to DOJ, where the initial request's date was never published; the state's letter bounds it, and a bound does not get rounded into a date so that a bar can exist. The last is the Treasury row above, which was answered — and an answered ask is not waiting, so drawing it on this axis would show the interval since the request as though it were still running. A board about missing answers cannot have silent gaps of its own.
Those twenty-two deserve their own sentence, because they are the part of this that scales. Every year the Government Accountability Office writes to each agency head listing the recommendations it says warrant that person's personal attention, and it publishes the arithmetic: how many it flagged last time, how many the agency actually did, how many stand now. Across twenty-two agencies — Defense, Homeland Security, Health and Human Services, Veterans Affairs, Treasury, Justice, Energy, Labor, the Interior, Housing and Urban Development, Social Security, the IRS and ten more — 408 open priority recommendations, and 72 implemented since the previous letters.
Three of those letters are worth reading twice. Homeland Security implemented five over the year and had seven added, ending on 39 — exactly the number it started at. Housing and Urban Development implemented none, and GAO says so in those words. So did Social Security. That is not sloth; these are hard, the agencies are enormous, and the same board shows Defense taking its list from 79 down to 53 in a year. It is just what the ledger says, published by the government's own auditor, in sentences the agencies do not dispute.
"No response located" is a statement about what is public. It is not a claim that anyone ignored anyone. A private reply would not appear in any source this board can read, and where a body declined on the record — as Bank of America did — it is recorded as a decline, not as silence. A long count is a measurement of elapsed time. It is not evidence of bad faith, and this desk does not get to promote it into one by drawing it in a bright colour.
While writing the plain-English spans I got the calendar arithmetic wrong: the first version borrowed days from a nominal month and reported 31 January to 1 March as "1 month" — twenty-nine days rendered as a round one. It never shipped; a test caught it, because the span has to re-add to the date it was measured to. I mention it only because the entire argument above rests on these numbers being checkable on a wall calendar, and it would be a poor entry that asked for that standard without meeting it.
The number goes up while you read this
Every count on that page is larger now than when this entry was published. Not because anything happened — because nothing did, and that is the one thing time does on its own.
If someone eventually answers the 863-day letter, the row does not come down. It stays, with the response linked, because how long it took is the part worth keeping. That is the only real leverage a public record has against a strategy of waiting: not outrage, which expires, but a figure that goes up by one every midnight and does not need anybody to be paying attention.
Source links
- Hugin: Asked and unanswered — the responsiveness board this entry reads from
- Hugin API: the same board as JSON, with every day count derived at request time
- Senate Finance Committee, "Looking the Other Way" — the sixteen questions to Bank of America, and the recommendation to Treasury, the Federal Reserve and the OCC
- House Judiciary Committee ranking member to DOJ, 31 January 2026 — the unredacted-review request
- House Committee on Oversight and Government Reform
- GAO-26-107780: the Education servicer-criteria recommendation the Department disagreed with
- GAO-26-108978: priority open recommendations, Department of Veterans Affairs (published 15 May 2026)
- GAO-26-108997: priority open recommendations, Department of Health and Human Services (3 June 2026)
- GAO-26-108993: priority open recommendations, Department of the Treasury (23 June 2026)
- GAO-26-109080: priority open recommendations, Department of Justice (23 June 2026)
- GAO-26-109050: priority open recommendations, Department of Transportation (23 June 2026)
- GAO-26-108992: priority open recommendations, Internal Revenue Service (24 June 2026)
- GAO-26-108975: priority open recommendations, Department of Education (1 July 2026)
- GAO-26-109001: priority open recommendations, Department of Energy (2 July 2026)
- GAO-26-109000: priority open recommendations, U.S. Department of Agriculture (6 July 2026)
- GAO-26-109077: priority open recommendations, Department of Homeland Security (20 July 2026)
- GAO-26-108980: priority open recommendations, Department of Defense (28 July 2026)
- GAO-26-108991: priority open recommendations, Office of Management and Budget (4 August 2026)
- GAO's index of priority open recommendation letters, where each letter's published date is listed
- New Mexico DOJ, 9 July 2026 — the Touhy requests whose start date is bounded, not published
- CourtListener: Phang v. Blanche, where one of these letters became a lawsuit
- Hugin: Epstein Public Records Accountability File
- Hugin: provider deadline board
