The federal weekly series printed $5.967 for the week ending September 7, 2026, the highest of its 1,695 readings and 15.7 cents above the June 2022 high. Run that 2022 high through the CPI-U and it is about $6.55 in July 2026 money, so in real dollars the older print is still the taller one. AAA's page read $6.1602 on the morning of September 12. Here is which number is which, and why that stops being pedantry the moment it reaches your grocery bill.

The version of this piece that went up on September 8 told you to mark September 9, because that is when the Energy Information Administration's next weekly diesel print was due, and because it might come in above the June 2022 high.

It did.

So half of the headline on this piece was wrong by Wednesday, September 9, and it was wrong in exactly the direction the piece itself predicted. The other half was not wrong, and it is still here. That is the correction, stated plainly and at the top where corrections belong, and it is worth being precise about which half is which.

What was false: the old sentence saying that no 2026 figure is a record on any series. In nominal dollars, on the EIA's weekly series, one now is.

What was not false, and is not being withdrawn: the real-dollar finding. The earlier version ran a named price index against the June 2022 peak and published the result, and that work stands. It has been re-verified at the source for this correction, and it is in the section called "And then there is the dollar," with the index, the two values and the arithmetic where you can check them.

What was promised was that the rule would not move when the number did, and the rule has not moved. Name the series. Name the dollar. The sentence still holds, and it just gained one clause.

Here is the clause, and then the three numbers now in circulation, all of which are correct and none of which mean the same thing.

What printed on the ninth

The EIA's weekly survey of U.S. No 2 diesel retail prices, series EMD_EPD2D_PTE_NUS_DPG, printed $5.967 a gallon for the week ending September 7, 2026. The agency released it on September 9.

That series is long. It carries 1,695 dated weekly observations going back to March 21, 1994, when a gallon of diesel averaged $1.106. Its previous high was $5.810, set the week ending June 20, 2022. So in nominal dollars $5.967 is the highest figure in thirty-two years of that survey, by 15.7 cents.

And here is the cleanest way to see how much of a nominal outlier it is: of all 1,695 readings ever published in that series, exactly one sits above $5.810, and it is this one.

That is a record. It is a nominal record, which is a word this piece is going to keep saying, and the reason is in the section after next.

Three numbers, and all three are right

This is where most coverage goes into the ditch, so we are going to walk it slowly.

The federal weekly survey says $5.967, for the week ending September 7. That is the EIA series above. It is a weekly average across a national sample, which means it is not a daily price and it is not the price at any particular pump, including yours.

AAA's own page said $6.1602 on the morning of September 12. The page is live and it stamps itself: "Price as of 9/12/26." Read that morning, AAA's national average table showed diesel at a "Current Avg." of $6.1602, against a "Yesterday Avg." of $6.0556, a "Week Ago Avg." of $5.8819 and a "Month Ago Avg." of $5.3575. AAA's page also keeps its own record row, and it now names $6.1602 on September 12, 2026 as the highest average it has ever recorded for diesel. Nothing on that page adjusts for inflation, so that is a nominal record too, on AAA's own series, by AAA's own reckoning.

And AAA's page put diesel at $3.7029 a year ago. Against the $6.1602 reading of September 12, that is a rise of roughly 66 percent in twelve months, which is our own division of two numbers printed in the same column of the same table. There is nothing soft about that, and you do not need a second survey to establish it.

Now the discipline, and it is the whole reason this piece exists. Those first two numbers are 19.3 cents apart and neither one is an error. Before anything else, they are not readings of the same stretch of time. The EIA figure is a weekly average for the period ending September 7. AAA's is a point reading taken five days later, and AAA's own table shows how much can move inside a window that size, because the "Week Ago Avg." on it sits 27.8 cents below the number it printed that morning. On top of the calendar, the two are different instruments: AAA runs its own retail survey with its own sample and its own method, the EIA runs a different survey with a different sample and a different method, and instruments disagree.

Nothing on either page tells you how much of the 19.3 cents is the five days and how much is the method, so this piece is not going to tell you either. The gap is the gap. The dates are on it.

What you do not do is split the difference. What you do not do is write one number and cite the other. What you do not do is call the gap a mistake by whichever agency you like less. You name the series every single time, and then nobody can take your sentence apart.

And then there is the dollar

Neither survey adjusts for inflation. The EIA publishes that weekly series in nominal dollars and carries no deflator alongside it, so a 2022 number and a 2026 number sit in the same column as though a dollar meant the same thing in both years. It did not.

So run the old peak through an index with a name on it. The Consumer Price Index for All Urban Consumers, series CUUR0000SA0, U.S. city average, all items, not seasonally adjusted, read 296.311 in June 2022 and 333.918 in July 2026. Put the June 2022 peak through that and $5.810 then is about $6.55 in July 2026 money. Measure the new print against that number instead of against the raw 2022 figure, and $5.967 comes in about 8.9 percent below it, which is 58 cents short.

That is the comparison the earlier version of this piece ran, and the September 9 release did not touch it. In July 2026 dollars the June 2022 observation is still the taller of the two, and the federal weekly series has not set a real-dollar record. What printed on the ninth is a nominal one.

Two limits on that, and both belong in the open. July 2026 is the vintage. August 2026 has since posted at 334.980, and running the same arithmetic on the August index gives about $6.57 instead of about $6.55, which moves the figure two cents and the finding not at all. And AAA's page carries no prior diesel record and no price index of any kind, so nothing here ranks AAA's series in real dollars. The real-dollar comparison above lives entirely inside the federal weekly series, which is the discipline of this whole piece: you do not measure one instrument against another one's arithmetic.

A real-dollar claim needs a deflator with a name, a vintage and a month. That is what CUUR0000SA0, July 2026, is. "Adjusted for inflation" on its own is not a citation, and anybody who hands you one without those three things has not done the work.

So the airable sentence, and you can carry it anywhere: this is a nominal record on the federal weekly series, and a nominal record on AAA's series, and in July 2026 dollars the June 2022 peak is still the higher figure on that federal series. Three clauses. All three true. Nothing in it can be corrected out from under you.

The $108 headline is about Brent

Crude is where the causation argument lives, and it splits the same way.

The EIA's daily spot series for Brent printed $109.51 a barrel on September 9. Its daily spot series for West Texas Intermediate printed $97.26 the same day.

So when you see a headline saying oil is above $108, that headline is true of Brent and false of WTI, on the same date, from the same agency. Twelve dollars a barrel sits between two numbers that both get called "the price of oil." Same lesson as the diesel numbers, same fix: say which one.

What the agency actually says, and what it does not

This is the part worth being careful about, because it is the part where a true story gets spent on a claim nobody can back.

The EIA's September Short-Term Energy Outlook does give reasons, in its own words, and the first thing to get right is which price each sentence is about, because the outlook labels them and most of the coverage does not.

Under the agency's own heading Global oil prices, it says those prices "remain elevated in response to falling global oil inventories, which we estimate have decreased by 400 million barrels so far this year." That is crude. The sentence right before it in the same paragraph is "Global oil prices rose to an average of $91 per barrel (b) in August, $7/b higher than in July," so there is no mistaking the subject.

Diesel gets its own paragraph, one heading down, under Distillate fuel oil inventories, and that is the one that matters here. It says "Tightness in the global distillate market has raised domestic prices and incentivized U.S. exporters to increase distillate exports." That one is stated flatly. In the next sentence it assumes global distillate production will stay below last year's levels, "contributing to low U.S. diesel inventories and high diesel prices," which is the agency putting its own name on the diesel mechanism while labelling the production figure as an assumption.

And under Global oil market assumptions, it builds in an assumption rather than a finding: "we assume some constraints to exporting oil from the Middle East will persist through the end of the year."

Read the distillate sentence again, because it is doing more work than the other two. The agency is saying the squeeze in diesel specifically pushed domestic prices up and made it more attractive to ship the stuff overseas. Tight market, higher price at home, and a stronger reason to export. That is a mechanism, it is about diesel, and it is the agency's own.

Now the thing that is not there. The word "Iran" does not appear in the text we read. Not once. The outlook refers to production sitting below "pre-conflict" levels without naming the conflict at all.

There is a war framing moving through the coverage, and it says the Iran war is what put diesel over six dollars. It may well be right. Inventories fall for reasons, shipping constraints in the Middle East have causes, and the agency's own "pre-conflict" language is plainly pointing at something. But an inference drawn from an agency's silence is an inference, not a finding, and it does not get to wear the agency's name. If you put it in the EIA's mouth, the first person who opens the outlook takes your credibility, and they will be entitled to it.

Say it the way it is. On diesel, the agency says tightness in the global distillate market "has raised domestic prices," and it assumes production below last year's levels will keep contributing to low inventories and high prices. On crude, it points at falling global oil inventories. On the Middle East export constraints, it says it assumes them, which is not the same as finding them. The war reading is a reasonable inference on top of all that, made by people other than the agency.

You are already paying it, and not at the pump

If you do not own a truck, you might have read this far thinking the number is somebody else's problem. It is not, and the reason is a distinction almost nobody makes out loud.

Diesel is not a consumer price. It is an input price.

Gasoline is a thing you buy. When it goes up, you feel it that afternoon, at the pump, in your own hand. That is why gasoline gets the headlines and the photo of the price sign.

Diesel is a thing that gets bought on your behalf, over and over, by people you will never meet, before anything reaches you. It moves the groceries to the store. It runs the freight, the concrete mixer, the school bus, the combine at harvest. It is inside the delivery fee, the invoice your boss reads before deciding what raises look like, and the few cents that show up on a box of cereal. Nobody ever sends you a bill with the word diesel printed on it.

Which means the nominal record the federal weekly series set for the week ending September 7 does not land in September. It lands in October, and it lands in November, in the price of things that have nothing obviously to do with fuel, and by the time it arrives it will have been renamed four times.

And that is precisely why the series question is not a pedantic aside. A reader who cannot tell which number is which cannot tell whether they are being told the truth about their own grocery bill. When the increase finally shows up on the shelf, somebody is going to explain it to you. Knowing which survey says what, and in which dollars, is the difference between checking that explanation and swallowing it.

The strongest version of the other side

State it before you answer it, and state it at full strength: who cares which series it is. Diesel is over six dollars either way. A family choosing between a fill-up and a grocery run is not helped by a lecture about survey methodology, and the hair-splitting is a way of looking rigorous while the actual harm goes unaddressed. Precision can be a hiding place.

That is fair and it deserves a real answer, not a dismissal. Here it is.

Precision is not the alternative to anger. It is what keeps the anger on its feet. The version of this weekend where the show yells "record diesel" with a number it cannot source spends the next week arguing about a chart instead of about who is getting paid, and it loses that argument, and the family choosing between the fill-up and the groceries gets nothing out of the exchange at all. Every correctable sentence is a free win handed to whoever would rather the subject changed.

The two things are not in tension. Diesel is up about 66 percent on AAA's series in a year. That fact gets stronger when it is nailed down, not weaker.

What you can actually do

  1. Before you repeat a fuel record, open the series and read the date. The EIA's weekly diesel series is published as a spreadsheet with the week-ending date on every row. Two clicks, and you never post a sentence somebody can take apart.
  2. Say which dollar out loud when you say "record." If nobody in the chain ran a named index, the word is nominal. It costs you one syllable and it buys you the whole argument.
  3. When somebody hands you the price of oil, ask which benchmark. Brent and WTI were twelve dollars apart on September 9. A headline built on one of them is not a claim about the other.
  4. Mark September 15, 2026. That is the EIA's next scheduled release for this series, so it is the date this number either holds or moves. The rule holds either way.
  5. When somebody hands you an all-time high, ask two questions. Which series, and which dollar. If they cannot answer both, they do not know whether their own number is true, and now you do.

The so-what

Three numbers went around in the week of September 7 and every one of them was correct. $5.967 on the federal weekly survey. $6.1602 on AAA's page on September 12. Oil above $108, if you meant Brent and not the benchmark that printed $97.26 on the same day. Nobody had to lie to produce that confusion. They only had to leave off which instrument they were reading, and in whose dollars.

That is worth noticing, because it is how most of what you get told about your own cost of living actually works. Not a fabricated number. A real number with its label filed off, arriving right when somebody needs you confused about whether things are getting more expensive and who arranged it.

The nominal record is in the ground now. The bill shows up in October, in the price of food that never mentions fuel. Being right about which number is which is not fussiness. It is the only thing that keeps you standing when the argument gets to the part that matters, which is who is getting paid.

Come back for the rest

We do this every week. If you want the documents while they still matter, The Long Pour is where they land. Pull up a chair, pour something cold, and let us send you the receipts before the next quiet robbery clears the courthouse.

Related from this desk: The Poverty Tax on what it costs to be broke in Florida, and The JTA Budget Hole on who actually rides when getting around gets expensive.

Y'all come back, now.

The Receipts

  • Weekly U.S. No 2 Diesel Retail Prices (U.S. Energy Information Administration, weekly series EMD_EPD2D_PTE_NUS_DPG, week ending September 7, 2026, released September 9, 2026; accessed September 12, 2026): the $5.967 print, the $5.810 June 2022 prior high, the 1,695 observations back to $1.106 in March 1994, and the September 15 next release.
  • AAA Gas Prices, national average (AAA, national average table, live page, its own stamp reading "Price as of 9/12/26"; accessed on the morning of September 12, 2026): the $6.1602 current, $6.0556 yesterday, $5.8819 week-ago, $5.3575 month-ago and $3.7029 year-ago diesel figures, and AAA's own record row.
  • Europe Brent Spot Price FOB (U.S. Energy Information Administration, daily spot series RBRTE, September 9, 2026; accessed September 12, 2026): the $109.51 Brent print.
  • Cushing, OK WTI Spot Price FOB (U.S. Energy Information Administration, daily spot series RWTC, September 9, 2026; accessed September 12, 2026): the $97.26 WTI print on the same date as the Brent figure.
  • Short-Term Energy Outlook (U.S. Energy Information Administration, September 2026, Forecast overview, headings Global oil prices, Distillate fuel oil inventories and Global oil market assumptions; accessed September 12, 2026): the three quoted sentences under those headings, and the absence of any reference to Iran.
  • Consumer Price Index for All Urban Consumers (U.S. Bureau of Labor Statistics, series CUUR0000SA0, U.S. city average, all items, not seasonally adjusted, June 2022 and July 2026; accessed September 12, 2026): 296.311 and 333.918, the two index values behind the $6.55 rendering of the 2022 peak.