The last daily price the United States government published for Brent crude is $130.80 a barrel, dated 15 September. The contract that actually sets the Brent price traded at $98.82 at half past ten last night.

Nothing official has said so yet, and nothing official will say so before this email reaches you. The EIA's daily Brent series stopped on 15 September. The EIA's own page says the next release is today, and nothing in its Wednesday petroleum cycle publishes before half past ten in the morning. On Monday we told you this issue would carry the first primary oil print since the 15th. It does not. That is a publication calendar, not a missing number, and saying it once is better than hedging it twice.

Here is what the exchanges do give us, and the useful figure is not the level.

ICE Brent for November was $98.82 at half past ten last night, down about 1.5% on the session on 422,611 lots. CME's WTI contract for the same month was $90.00 at the same hour. Both boards run on a delay of at least ten minutes, and nine dollars sits between them. On the EIA's last published pair, 15 September, world crude sat $23.78 above American crude. On 1 September the gap was $4.54. It more than quintupled in a fortnight and has now closed most of the way back.

Why that particular gap and not the price: Brent is seaborne and WTI is landlocked at Cushing, Oklahoma. Anything that interferes with getting barrels onto water moves one and leaves the other where it is. A blowout in the spread is the market pricing a problem with shipping. A collapse in the spread is the market pricing that problem receding. It says nothing about who did what, and we are not going to tell you whose ships or whose strait, because we could not confirm any of that from a primary source in time to print it.

The last settlement session CME had published when we wrote this was Monday's, and Tuesday's will be on that page by the time you read this. For trade date 21 September the November contract settled at $92.37, down $3.71 on the day, on estimated volume across the curve of 837,729 lots. The October contract expired on Tuesday, so November is the front month you will find on that page this morning.

Now the part that has not moved.

The EIA's weekly gasoline and diesel update, out yesterday, has regular unleaded at $4.478 a gallon, up 15.9 cents in one week. Diesel is $6.529, up 24.4 cents in a week and 74% higher than a year ago. Retail fuel lags crude by weeks, so the cause is receding while the cost is still arriving. The Midwest took the worst of it rather than the coasts: Chicago added 45.9 cents in seven days. Denver is the only line in the entire table that fell.

Which brings us back to a sentence we made a point of on Monday. July's statement blamed inflation in part on "supply shocks that have driven price increases in certain sectors, including energy." September's statement does not contain that clause. We wrote that the committee deleted it in the week energy would most have justified keeping it, and Brent's highest print of the whole episode was indeed the day before the decision.

That reads differently this morning. With crude back near $99 the deletion looks early rather than vindicated. It could be either. A committee that removes its energy excuse a week before the excuse expires has either seen something or has simply written the paperwork a rate rise requires, and there is no way to tell from outside which one happened. We are not going to pretend the second reading is the obvious one just because we published it first.

What we do with the EIA number when it lands: Monday's issue gets it, two weeks of official spot in one go, printed against the futures curve above. If the government series comes in materially away from the exchange tape we will show both and say so.

Three metrics.

Metric 1, Fear and Greed: 71, Greed. The stamp is taken at midnight UTC, so this is last night here. Tuesday's reading was 78, Extreme Greed. On the evening of the decision itself it was 50, Neutral. Twenty eight points in six days, every one of them landing after the hike rather than before it, and seven already given back.

Metric 2, ETF flows: plus $999.0 million on Monday, with no fund in outflow. We checked that against Farside's full published history rather than take anyone's word for it. There are 692 daily rows on that page going back to 11 January 2024, and the last one bigger than Monday is 6 October 2025 at plus $1,205.2 million. Eleven and a half months. Monday's was IBIT $381.4 million, ARKB $289.1 million, Fidelity's FBTC $238.8 million. As we write, Tuesday's row reads plus $364.4 million and it is incomplete: IBIT, the largest fund in the complex, had not reported, and Farside prints a dash rather than a zero where its number belongs. Treat Tuesday as provisional and look at the row again before you use it. September stands at about plus $1.31 billion through Monday, which is our own sum of Farside's published daily totals rather than a figure Farside prints, and SoSoValue's published monthly number agrees to the rounding. One name to learn if you follow that table: MSBT, which only appeared on it on 8 April and was the second largest contributor on Tuesday at $99.0 million.

Metric 3, BTC dominance: 58.77%. Up about seven tenths of a point from the 58.1% we printed Monday, on a total crypto market of $2.95 trillion late last night. The corporate line has new information rather than a correction. Monday's issue described the disclosure covering 8 to 13 September, which showed no bitcoin bought. A second filing, accepted half an hour after Monday's email went out, covers the following week, 14 to 20 September: 950 coins for $75.7 million at an average of $79,670, taking the pile to 846,000. In the same seven days it spent $174.0 million buying back its own STRC preferred. Both out of cash, with no shares sold under the at-the-market programme to fund either. It put 2.30 times as much money into its own stock as into bitcoin and raised nothing to do it. MSTR closed Tuesday at $167.33, down 0.69%, after 9.47% on Monday. From the close on decision day to Tuesday's it is up 32.6%, against bitcoin's 13.2% over the same four sessions.

Nothing grades in the STACK today, and we would rather write that than manufacture something. The book is flat for the fifteenth issue running and the live levels on all five open tests are down there.

📣 THIS SPOT IS OPEN

Baseline Crypto reaches nearly 7,000 Bitcoin-focused readers every Monday and Wednesday morning, opening about half of what we send. Direct sponsor placements are open, premium position, top of the issue, one per send.

Reply to this email for rates.

A statement clause is not a forecast. It is a permission slip, and what the committee tore up was the one that said this inflation was never its problem to fix.

Take the charitable reading seriously first, because the timing really is striking. July's statement carried the supply shock language. September's does not. A week after it came out, the front Brent future is under $100 and the spread that measured the disruption has closed from $23.78 back to about nine dollars. The committee also had the futures curve in front of it on the day, and that curve was already in steep backwardation, which is a market saying the tightness is temporary. Nobody who looks at this and concludes the Fed read it correctly is being stupid.

Now ask what the clause was for.

A central bank is paid for one number. The supply shock sentence is the only line a policy statement ever contains that says out loud: this inflation is not the kind interest rates fix. In July that sentence was worth having. It is the justification for sitting still while three of your colleagues vote against you for a hike. By September, having decided to raise, the same sentence turns into a liability, because you cannot credibly raise rates to solve a problem you have just finished telling everybody rates do not solve. The clause did not leave because twelve people formed a view on tanker traffic. It left because it contradicted the action.

The cost of that is ownership, and it runs one way. Diesel is 74% dearer than it was a year ago and it rose 24 cents last week alone. If the pump keeps climbing through October, supply shock is no longer available as a description, because they deleted it themselves. If the pump falls in November because crude fell in September, they do not get the credit either, because the mechanism that would have connected the two is the same clause. That is not a small thing to give up in exchange for tidier paperwork.

Which makes the silence since worth noticing. The two people who have said anything substantive about this hike do not vote on the next one. Tom Barkin of the Richmond Fed gave a speech on Tuesday called Why Hike?, and the entire forward guidance in it is this: "Will additional hikes be required, and how many? We'll see." His actual argument is about breadth rather than energy, that more than 60% of the PCE index is now rising faster than 3% a year, which is the opposite of a supply shock story. Austan Goolsbee spent Monday in London arguing that central banks should stop looking through supply shocks when the shocks are persistent. Neither man has a vote in 2026.

Of the twelve who do, three appeared in public during the window and none said a word about rates: Jefferson on discount window plumbing and Williams on market structure, both at the same New York Fed conference, and Bowman twice in London, on supervision. Hammack, Kashkari and Logan, the three who dissented in July for precisely this increase and then got it, have published nothing since the decision.

So the market is pricing October at a coin flip on the testimony of two non-voters and a deleted sentence. The next voting member to speak is Barr, at five past ten this morning on housing, and then Hammack at ten to nine on Thursday. Both are after this email. Until one of them says something, the loudest voices on this hike belong to people who cannot vote for the next one.

Three numbers that define the next week.

$2,921.5 billion, which is where the series we bet on actually ended the week. On Monday we printed reserve balances at $3,013.8 billion and put a dated test on the week ending 23 September. Both of those are correct and they do not sit together the way we implied. WRESBAL, the series our test names, is a weekly average. The week ending 16 September opened from a Wednesday level of $3,036.5 billion and closed, after the 15 September tax date, at $2,921.5 billion on the Wednesday level series. The average was flattered by the days before the drain. The gap between the two is $92.3 billion, and that gap flips sign in half the weeks we pulled, so treat it as an artefact of comparing an average with a snapshot rather than as a drain signal.

The week being graded runs Thursday 17 September through Wednesday 23 September, and it starts from the lower figure. So the arithmetic, which anyone can check: for the print to reach Branch C, $3,000 billion or above, reserves have to average about $78.5 billion higher than the level they closed at on the 16th, across the whole week. Over those same days the Treasury took another $31.7 billion out. Its general account closed at $972.675 billion on the 17th and $1,004.391 billion on Monday, the first close above a trillion since 31 August. The facility that used to soak up drains like this took just $453 million on Tuesday, against $5.4 billion as recently as the 16th.

We are not grading it today and we are not pre-grading it either. The test resolves in Monday's issue on the Federal Reserve's published number, not on our estimate, and the H.4.1 that carries it does not come out until Thursday at half past four. What we will say, because we published the position and it is moving underneath us: Branch C now requires a large average increase in a week that drained.

One satoshi per byte, on every tier. mempool.space's recommended fee returns 1 sat/vB for fastest, half hour, hour and economy, all four at the floor, while the price is up 13% in six days. That combination does not usually last. The 600 coin transaction in the RADAR above cost 5,744 satoshis, roughly five dollars, to move about $52 million. What to do with it is the thing you have been postponing because the fee felt like a reason: consolidating scattered UTXOs, getting coins off an exchange, rotating to a new wallet after a firmware change. None of that gets cheaper than this.

2.66%, four basis points inside the line we drew for ourselves. On 2 September we committed in print that if high yield spreads were sitting at or inside 2.70% when the Treasury's cash balance peaks in late October, our framing of this whole liquidity cycle was wrong. The ICE BofA high yield spread printed 2.76% on 15 September going into the meeting, 2.70% on the day of it, and 2.66% on Monday. Credit has now watched a first rate rise in three years, a government move $120 billion into its own account in a single day and an oil price that ran 25% in six sessions, and it has tightened through all of it. Our 7 October test has this in Branch C. The commitment has about five weeks left to run and the number is moving against us the whole time, but it is not due yet and we are not going to claim it early. If the spread is still at or inside 2.70% when the cash balance peaks, the sentence we owe you is that our framing was wrong, in those words.

Forget analyst predictions. Polymarket is a real-money prediction market, where traders put actual dollars on outcomes. Scorecard first, then the board. Not financial advice. Our read. Disclosure: we hold personal positions in Polymarket itself and may earn a commission from Polymarket referrals. These markets are thin and the odds below move, so read direction over ticks.

Scorecard. Nothing grades today. No dated test resolves on 23 September. We are not going to invent one so that this paragraph has a verdict in it. The next is reserve balances for the week ending 23 September, which grades in Monday's issue on Thursday's H.4.1, and four more take their readings between 30 September and 7 October. Here is where all five actually stand this morning.

Reserve balances, week ending 23 September, grading Monday 28 September. Branch A below $2,900 billion, Branch B $2,900 billion to under $3,000 billion, Branch C $3,000 billion or above. Last published point is the week ending 16 September at $3,013.8 billion. The arithmetic that has changed since we set it is in BEYOND THE CHARTS above, and the short version is that the week we are grading opened $92.3 billion below the number we printed.

The Treasury cash close on 30 September, grading 5 October. Branch A $1,050 billion or above, Branch B between $950 billion and $1,050 billion, Branch C below $950 billion. Live: $1,004.391 billion on Monday's Daily Treasury Statement, Branch B, having crossed a trillion on the way up. Tuesday's figure publishes this afternoon, after this email.

US spot bitcoin ETFs net positive for September, grading 7 October. Live: about plus $1.31 billion through Monday, Branch A. Six sessions left, and Farside's own history contains single days worse than minus $900 million, so that cushion, with IBIT's Tuesday number still missing and likely to add a few hundred million more to it, is not a result.

High yield spreads on 30 September, grading 7 October. Branch C is 2.40% to under 2.70%. Live: 2.66% on Monday's print, inside Branch C and inside the separate 2 September commitment that says our framing is wrong.

The ten year real yield on 30 September, grading 7 October. Branch B is 2.55% to under 2.80%. Live: 2.62% on Monday. The last five sessions read 2.62, 2.68, 2.61, 2.68, 2.62, which is a series with no trend and a lot of noise, and one week from the measurement date it is still sitting mid branch.

The book. Flat for the fifteenth consecutive issue, recomputed from the archive tonight rather than carried over from Monday's number. The rule we wrote down on Monday still governs: a call counts as a win or a loss only when a published issue printed its exit, or the market resolved while we held it and an issue said so. By that rule nothing is open, and the last thing to go into the win or loss columns was an exit we printed on 3 August. Fourteen was too many when we said so on Monday, and fifteen is worse, so we will not say it twice. What we are waiting for has not changed either: a price one of our own published rules actually reaches, on a market whose resolution language we understand, with enough time left that being wrong is cheap. Two rules are still live. High yield spreads through 3.00% into the late October cash peak, which went further the wrong way this week. And the October leg below 40%, which is the first entry below.

Fed decision in October, 25 bps increase | Market: 52.5% 🟡 NO TRADE, and Monday's rule stands unchanged
No change is 45.5%. Everything else on that board, cuts of either size plus a 50 basis point increase, adds to under 1.6% combined. The path since Monday is worth more than the level: it was 54.5% at midnight and 49.5% at the equity open, fell to 48.5% by twenty to twelve that morning, was back at 55.5% before dawn on Tuesday and sits at 52.5% now. It round tripped and spent part of Monday below even money, in a week when not one voting member said anything about rates. What would put us in: this leg below 40%. One number, one market, no second clause.

Fed decision in December, 25 bps increase | Market: 67.5% 🟡 NO TRADE, and this is where the October argument actually settles
No change in December is 27.5%. Put the two meetings side by side and the market is not arguing about whether there is another increase this year. Another hike in 2026 trades at 86.5%, and on the market that counts them, which includes the September increase already delivered, two hikes is 63% and three is 24%. So the argument is only about which of the two remaining meetings it lands in, with roughly a one in four chance on both, and that is the same thing the committee's own median does when it projects one more increase and declines to name the date. What moves this leg is October itself: a hold on 28 October lifts it in a single afternoon and a hike collapses it. We are not paying 67.5 cents to sit for five weeks waiting for an event we can watch for free.

Senate passes the Clarity Act by 25 September | Market: 0.15% ⚪ NO TRADE, and the number is the point rather than the price
Fifteen basis points, eight days after cloture failed 49 to 50, eleven short of the sixty it needed, on the day the Senate came back and spent its afternoon elsewhere. The leg carries $7,188 of volume against a one tenth of a cent bid and a two tenths of a cent ask, so this is not a price anybody is defending. It is the absence of one. The later legs on the same event read 1.15% for 2 October, 1.80% for the 9th and 4.50% for the 31st, while the parent contract on the bill being signed this year fell from a high of 7.65% on Monday morning to 5.95% now. Nothing here is tradeable at these sizes. It is a thermometer, and it reads cold.

Zero Fed rate cuts in 2026 | Market: 96.45% ⚫ EXITED at 88.75%, a loss
We were short from 78%, trimmed at 85% and covered the rest at 88.75% on a published stop on 3 August. It is 96.45% now, so the loss kept getting worse after we took it, which is the ordinary fate of a stop taken on a short that was wrong in the first place. Event volume is $53.2 million, far and away the deepest market on this board. What would move it back is not a soft data print and not a speech. It is one dot below the current midpoint in the December projections, from anybody.

Track everything live at polymarket.com, free, no account required.

On Monday we asked you one question: does the Fed raise again on 28 October, yes or no. We said we would print your split next to the market's.

There is no split. We went through the inbox and our own file of reader replies before writing this, and there has been nothing new since 3 August. That is three issues in a row with none.

Three is enough to stop blaming the week. The question is the problem, and it is our question, so this is on us. We keep asking you to forecast, which means doing homework and then defending a position to somebody you have never met, at half past seven in the morning. Almost nobody wants that. So we are changing what we ask for. From today it is one word or one number, something you already know without looking anything up, and we print the distribution rather than the argument.

One number: what does a gallon of regular cost at the station nearest you today? Reply with the number and nothing else.

The national average is $4.478 and the spread hiding inside it is enormous. Chicago went up 45.9 cents in seven days to $4.879. California is at $6.003. Denver is the only line in the EIA's table that fell. If enough of you send a number, Monday's issue prints what our readers are actually paying against the government's average, and we do not know of anyone else who publishes that.

Hit reply. We read every response, and the best of them run Monday.

See you Monday with two weeks of official oil prices in one go, Thursday's reserve balance print and the grade on the test attached to it, and whatever the first voting members to speak since the hike decide to say.

The Baseline Crypto Team

Two tools. Free, no account, nothing to install, and we checked both of them again last night.

The hardware wallet checker takes your device and firmware version and tells you whether you are inside a population with a known problem, and what to do about it if you are. The custody audit is nine questions about what happens to your coins when something goes wrong, and most people fail it on backups and inheritance long before the hardware is ever the issue.

This week the instruction writes itself, and it is for one particular person: whoever you know who still keeps coins on an exchange because moving them is a hassle. BitMEX shut at four this morning UTC. The 1% annual fee it now charges on balances nobody withdrew is not aimed at the traders who were watching the settlement. It lands on the person who has not logged in since last year and will find out when they eventually do. Network fees are at one satoshi per byte today, which is as cheap as this ever gets.

Do not send them a lecture about self custody. Send them the custody audit, let the nine questions do it, and then ask where their coins actually are. They will not enjoy the answer, which is the entire reason it was worth sending.