The Backroom Brief

Press Digest

The divergence is regional and it is about which shock each desk thinks is running. The American and British broadsheets lead with rates: the Wall Street Journal on the ten-year approaching five percent, CNBC calling the hike certain, Reuters writing that the hike case builds. The Gulf and Asian press lead with the barrel and the shipping lane. Business Recorder frames a rising Dubai market as a function of oil supply fear, which is a trade the American coverage does not describe at all.

The one story only one region is leading with is Japanese tightening. Nikkei Asia and the Reuters poll carriers have the Bank of Japan raising to 1.25 percent as effectively settled, while the American front pages treat the week as a single-central-bank event. Three of the four decisions this week are not American.

What Moved and Why

1. The hike stopped being a forecast and the forecasters did not move with it

Two prints did the work. August producer prices landed Thursday at 5.4 percent against a 5.1 percent consensus, and the thirty-year Treasury touched 5.35 percent, a nineteen-year high. Friday's consumer price index accelerated to 3.4 percent with core at 0.3 percent on the month, both above forecast (Bureau of Labor Statistics; Reuters and Morningstar, 11 September).

The repricing was violent and it is documented day by day. Traders had 70 percent on a September hike on 3 September. By Friday CoinDesk put it near 70, by Saturday Polymarket was at 83 and CME-derived buckets were at 86.5 (CoinGape, 12 September). As of this morning the futures strip shows 14.5 percent on a hold at 3.50 to 3.75 percent and 85.5 percent on a move to 3.75 to 4.00 percent (CME-derived buckets via Investing.com, updated 12 September). CNBC's markets desk wrote on Friday that a hike seems certain.

The stake is what the strip has stopped pricing. At 85.5 percent the market has taken Wednesday off the table as a question, which means every remaining uncertainty has moved into the projections rather than the decision. A Bloomberg survey cited secondhand put fewer than 13 percent of forecasters expecting any change; this desk could not source that survey directly, flags it as unverified, and is therefore not building an argument on it. The verifiable point stands without it: a 14.5 percent hold is not a small tail on an instrument this leveraged to the front end.

Watch. Wednesday's decision and, more informative, the Summary of Economic Projections attached to it. This is a projection meeting. If the Committee hikes and the dot plot still shows cuts in 2027, the market has the level right and the path wrong. A hold on Wednesday reprices the entire front end by more than the twenty basis points it just added.

2. The curve bear-flattened on real rates, and that answers less than it looks

The 09-10 issue argued the oil shock had acquired a monetary transmission channel and then admitted, in the same breath, that the cleanest evidence for it ran only through Wednesday. That evidence now exists.

Across Wednesday to Friday the two-year rose 20 basis points to 4.63 percent and the thirty-year rose seven to 5.35 percent, falling two basis points on Friday alone (US Treasury daily yield curve). Two-year to ten-year flattened from plus 40 to plus 33. Two-year to thirty-year flattened from plus 85 to plus 72. Twenty-year to thirty-year is inverted at minus three. The decomposition is the point: the ten-year real yield rose nine basis points to 2.55 percent on Thursday while the ten-year breakeven fell four to 2.36 percent on Friday, and 2.55 plus 2.40 reconciles to Thursday's 4.95 percent nominal exactly (FRED series DFII10 and T10YIE).

That rules out one explanation and only one. A flat real yield against a rising breakeven would have been simple energy pass-through, the inflation-compensation reading, and this desk set that test in advance on 2 September. The test came back negative, so the move is not the market marking up expected inflation because oil went up.

What it does not do is tell you why, and there are three candidates rather than two. A credible central bank gives this print. So does an economy whose growth is being marked down, though that normally pulls real yields lower rather than higher, so it does not fit this week. The third fits best: the producer and consumer price beats that triggered the repricing were themselves partly energy, which means the hike being priced is the oil shock transmitting through the Federal Reserve rather than a separate force competing with it. The prior issue treated the first candidate as established. It is not, and a four basis point move in a series that regularly travels that far in a day is thin evidence on which to have decided.

Watch. The two-year at 4.63 percent already sits 100 basis points above the 3.63 percent effective funds rate, so four quarter-point moves are in the bill. October buckets put 43.2 percent on a second hike to 4.00 to 4.25 percent. If Wednesday delivers and the two-year does not extend, the strip is done pricing and the flattening stops.

And note the transatlantic split this brief has so far named without reconciling. The contained-inflation leg above rests on a United States breakeven. The European Central Bank's own September staff projection has core inflation rising from 2.5 percent in 2026 to 2.6 in 2027. Frankfurt does not share the reading that the inflation problem is being contained on the current path, and it has the better claim to know its own economy. Contained inflation may be a purely American conclusion.

3. Four central banks, nine days, and only one of them is being watched

The European Central Bank raised all three key rates 25 basis points on Thursday 10 September, effective 16 September: deposit facility 2.25 to 2.50 percent, main refinancing 2.40 to 2.65, marginal lending 2.65 to 2.90. The decision was unanimous and Lagarde called it a no-brainer in the Q and A transcript. The press release reason was blunt: the Middle East conflict continues to generate inflation pressures and inflation will stay well above target for an extended period.

Two details the coverage is missing. This was the second ECB hike of 2026, not the first; the key-rates table carries an earlier move effective 17 June. And the September staff projections put headline inflation at 3.0 percent for 2026 and 2.5 for 2027, with core at 2.5 and then rising to 2.6. A core forecast that goes up in the second year is the Council telling you the June move did not finish the job.

Then the week ahead. The Federal Reserve decides Wednesday. The Bank of England decides Thursday, with Bank Rate at 3.75 percent since 30 July. The Bank of Japan decides Friday, with the overnight call rate guideline at around 1.0 percent and the meeting confirmed for 17 and 18 September at the Bank's own site. Reuters reported Sunday that speculators have turned net long yen for the first time since February.

Watch. The yen. A long-yen position built ahead of a widely expected hike is the most crowded expression of this entire week, and it is the one that pays least if the Bank of Japan delivers exactly what is priced. The asymmetry sits with a hold.

4. Saudi Arabia shut the pipeline that exists precisely for this situation

The East-West pipeline runs Saudi crude to the Red Sea so it does not have to transit Hormuz. Drones hit multiple pumping stations on Thursday and the shutdown was announced Friday evening (Financial Times, Reuters and Bloomberg, all on the evening of 11 September). Maritime Executive confirmed clear and substantial impacts at two pumping stations. Capacity is roughly 7 million barrels a day on Reuters and Al Arabiya reporting from earlier this year, though Business Insider Africa put it at up to 5 million on 12 September, and this desk prints the range rather than choosing. It is the second strike on this line in 2026; Iran hit it on 8 April.

The bypass mattered because the strait it bypasses is already shut. Hormuz transits ran seven vessels on Thursday against a ten-day average near fifteen (Reuters, 10 September; gCaptain). Meanwhile the Houthis took Mocha on Thursday and reached Perim island at the mouth of the Bab al-Mandab on Friday, halving traffic there to about 26 vessels a day (PBS, Guardian, Anadolu, 11 September).

The physical consequences are already printed. Saudi told OPEC that August crude output fell 1.9 million barrels a day to 6.238 million, the lowest since 1990, with exports at 3.2 million, a thirteen-year low. American diesel passed six dollars a gallon for the first time on record with refineries running at 98 percent, and the White House is weighing the Defense Production Act (oilprice.com and Reuters, 11 September). The International Energy Agency cut its 2026 supply outlook by 1.4 million barrels a day.

Watch. Whether the East-West line restarts this week. One bypass at 5 to 7 million barrels a day is the difference between a blocked strait being an expensive inconvenience and a blocked strait being a global shortage.

5. The weekend got worse and nothing could trade on it

Friday's settles are the last prices that exist. Brent closed at 103.51, down 3.8 percent, and West Texas at 98.85, down 3.5 percent (World Oil, 11 September). Both fell.

Then the news arrived. The Saudi pipeline shutdown was announced after the European close. The Financial Times reported Saturday that the United States is now limiting air-defence time slots for tankers transiting Hormuz, which is a rationing mechanism dressed as a schedule. Reuters led Sunday with a new report of an attack on Hormuz shipping. Bloomberg and the New York Times both ran the Saudi pipeline story Saturday, with Bloomberg framing the consequence as pressure on Iraq.

So the tape closed on a decline and the physical picture then deteriorated across two days in which no futures market could express it. The weekly numbers already disagree with the daily ones: Friday was down 3.8 percent inside the largest weekly gain since July, which Reuters put near 8 percent. Anyone reading Friday's close as the market relaxing about oil has the sign of the week backwards.

Watch. The Monday Asian open in Brent. That single print is the market's first opportunity to price the pipeline, the air-defence rationing and the Sunday attack report together, and it lands hours before the Fed's blackout ends.

6. The auction tape is inconsistent with a buyers' strike, with three caveats

The convenient story about a thirty-year at 5.35 percent is that nobody wants the paper. The September auctions say the opposite, and the arithmetic is checkable.

Treasury reopened the thirty-year on 10 September, 22 billion dollars, clearing at a high yield of 5.308 percent with a bid-to-cover of 2.61. Computed from the raw dollar awards rather than a vendor's percentage, indirect bidders took 17.453 billion of 21.957 billion competitively accepted, which is 79.48 percent, and primary dealers were left with just 2.21 percent. The ten-year reopening the day before cleared at 4.834 percent with indirects at 79.18 percent and dealers at 4.31 percent (TreasuryDirect auction records, CUSIPs 912810UW6 and 91282CRF0).

Dealer takedown of 2.21 percent is the lowest of any coupon auction in the last sixty days by a wide margin; the next lowest is 8.60 percent at the August ten-year. Dealers take down what end users refuse. A 2.21 percent residual means end users did not refuse. Add that the Federal Reserve's own add-on was zero at every September auction after running between 2 and 14.7 billion dollars through July and August, and the demand was genuinely private.

Three caveats, because this is the brief's boldest non-consensus claim and it should carry its own objections. First, both September prints were reopenings of existing securities, which mechanically attract index-replication and cash-futures basis demand that a new issue does not, so some of that bid arrives regardless of sentiment. Second, indirect is a bidding channel, not a buyer type. It captures foreign official accounts and real money, but also leveraged funds bidding through non-dealer intermediaries while financing the position in dealer repo, which is dealer risk one step removed and unwinds quickly under stress. Third, Treasury discloses no bidder identity, so 79.48 percent could be one concentrated account. This desk applied exactly that single-allocator caution to a 216 million dollar ether print elsewhere in this issue and owes the same caution here.

So the claim is downgraded from proof to evidence: the auction tape is inconsistent with a buyers' strike rather than proof there is none. The load-bearing support for this being a rates event rather than a credit event is not the auction at all. It is the independent option-adjusted spread data, where investment grade, high yield and emerging-market corporates all tightened in the same week.

Watch. The next thirty-year on the October refunding cycle. If yields keep rising while indirect awards stay near 79 percent, the long end is repricing policy, not credit risk, and the fiscal panic framing stays wrong.

7. Credit and short-dated volatility both refused to confirm, which is the trade

Everything above describes a market bracing for four central banks and a deepening energy shock. The two places that price stress most honestly did not agree.

Investment grade option-adjusted spread sits at 80 basis points, high yield at 270, and emerging-market corporates at 131, and all three tightened into the selloff (FRED series BAMLC0A0CM, BAMLH0A0HYM2 and BAMLEMCBPIOAS, dated 10 September, which carries a one to two day lag this desk states rather than hides). Volatility is the more interesting one, and it needs both timeframes to be honest. On the week the VIX rose 9.02 percent and nine-day volatility rose 20.89 percent, so the week did reprice risk. But on Friday, the session that carried the hot consumer price print, the VIX fell from 17.84 to 15.84 and nine-day volatility, the tenor that actually spans Wednesday's decision, fell from 17.70 to 14.47. It is now the cheapest point on the curve (CBOE daily price files). The market spent four days getting nervous and then sold the hedge on the day the data confirmed the fear.

The stake is simple. The market has agreed on the hike and priced none of the consequences. Selling the volatility that covers a projection-meeting FOMC, a Bank of England decision and a Bank of Japan decision inside 72 hours is a position that only works if all four banks do exactly what is expected and say nothing surprising. The 09-10 issue's reviewers warned that credit confirming rather than fading would turn a contained repricing into correlated deleveraging. Credit faded. That kept it contained, and it also removed the cushion.

Watch. Nine-day volatility on Monday and Tuesday. If it stays under 15 into the decision, the consensus trap is fully set. A move back above 18 before Wednesday means someone with size stopped believing the script. And note what equities did on the same Friday: the S and P rose 0.86 percent, the Dow 0.98 and the Russell 0.45, all into a hotter-than-forecast inflation print, while finishing the week down 0.80, 1.57 and 2.41 percent respectively. The small-cap underperformance is the only part of the equity tape behaving like rates matter.

8. The exchange-traded funds rotated out of bitcoin and into ether in the same week

American spot bitcoin funds took net outflows on every single session of the week: 46.65 million dollars out on Tuesday, 120.24 million Wednesday, 282.56 million Thursday and 13.29 million Friday, a four-session total of 462.73 million dollars with zero inflow days. That ends a three-week inflow streak. Net assets closed at 97.58 billion against holdings of 1,261,711 bitcoin (SoSoValue current metrics and historical series, both dated 11 September).

Ether funds did the reverse. Friday alone brought 216.41 million dollars in, of which BlackRock's ETHA took 148.82 million and its second vehicle another 18.32 million, against total ether fund assets of 16.31 billion. Same day, opposite direction, and the per-fund awards reconcile to the daily total to the cent.

The price followed and then gave it back. Ether pushed past 2,600 dollars for the first time in eight months on Friday, outrunning bitcoin, XRP and Solana, and trapped short sellers in a liquidation wave reported between 427 and 668 million dollars (Stocktwits and Yahoo Finance, 11 September; Bloomberg on the same mechanic). It now trades at 2,481.92, so the entire move round-tripped over the weekend. Bitcoin sits at 76,732, roughly 6.7 percent below the 82,284 high set the previous week.

Two endpoints of the same data provider disagree on cumulative bitcoin inflow by 21.4 million dollars, 55.176 billion against 55.155 billion. This desk prints both rather than picking the flattering one.

Watch. Whether ether fund inflows survive a down week. A single 216 million dollar Friday during a bitcoin outflow streak is either the beginning of a rotation or one allocator rebalancing. Two more sessions settle it.

9. Retail added into the fall while the people who trade size stepped back

This is where an earlier draft of this brief made a mistake worth owning. It paired bitcoin's retail ratio with ether and Solana's professional ratio and called the result a retail-versus-professional divergence. Checked within each asset, that divergence does not exist in bitcoin at all. Bitcoin's retail long-to-short account ratio rose from 1.2952 on Wednesday to 1.6539, and its top-trader position ratio rose too, from 2.1312 to 2.2324. Retail and professionals got longer together in bitcoin.

The real divergence is in ether and Solana, and it is stark. Ether retail went from 2.3102 to 2.7779 while ether top traders cut from 1.6463 to 1.2695. Solana retail went from 2.0731 to 2.3146 while its top traders cut from 2.4591 to 2.0471. So the people who trade size are reducing exactly where the fund flows are arriving, which is ether, and are not reducing in bitcoin, where the funds are leaving. That is a more interesting fact than the one the draft claimed, and it points the cascade risk at ether rather than bitcoin.

The venues disagree too, on two assets and not three. Hyperliquid funding is negative across every major, annualising at minus 12.21 percent on bitcoin, minus 4.28 on ether, minus 12.90 on Solana and minus 22.28 on XRP, so shorts pay longs there. Binance is mildly positive on bitcoin at plus 2.85 percent annualised and on ether at plus 3.53. Solana is negative on both venues, minus 4.76 on Binance against minus 12.90 on Hyperliquid, so the cross-venue sign flip covers bitcoin and ether only, not the whole board. One caution on the Hyperliquid alt book: a long tail of smaller names prints funding at exactly 10.95 percent annualised, which is that venue's structural floor, 0.01 percent per eight-hour period times three periods times 365 days, and it appears whenever the premium component clamps to zero on a thin book. It is never positioning.

Aggregate futures open interest fell to 59.5 billion dollars from 62.4 billion, and Binance bitcoin open interest fell from 8.362 to 7.980 billion across the week. Liquidations ran 256.3 million in 24 hours against 142.3 million midweek, and more than 140,000 traders were liquidated in Thursday's four-day drop (CoinDesk and TradingKey, 10 and 11 September).

Watch. Whether the retail ratio keeps climbing through Wednesday. Retail lengthening into a hike, with professionals cutting and open interest falling, is the configuration that produces the cascade rather than the bounce.

10. The venue this brief quotes is under investigation by its own regulator

On 11 September the Commodity Futures Trading Commission authorised three previously unreported insider-trading investigations into Polymarket, covering markets on Biden pardons, Iran and Google contracts. These are investigations, not charges, and no traders are named.

This brief quotes Polymarket prices in its lede and its prediction-markets section, so the disclosure belongs in the open rather than in a footnote. It also sharpens a live methodological question, because this issue leans on prediction-market pricing for the single biggest call in it.

The calibration held up well this week, which is worth saying because it usually does not. Polymarket's raw price for a 25 basis point September hike is 78.5 percent on 28.87 million dollars of volume. Applying the documented correction gives 84.5 percent. CME-derived futures buckets independently print 85.5 percent. A seven-point raw gap between two venues closes to one point after calibration, which is the formula doing exactly what it was fitted to do.

Watch. Whether the CFTC names traders. A prediction venue with a documented insider-trading problem in its geopolitical markets is a venue whose Iran and Hormuz prices this desk should stop treating as clean crowd information, and those are precisely the markets this brief has been citing.

Crypto

Bitcoin trades at 76,732 dollars, down 0.76 percent in 24 hours and roughly 6.7 percent below the 82,284 high set the previous week. Ether is 2,481.92, Solana 99.80. Total market capitalisation is 2.612 trillion dollars, down 4.51 percent on the day against 24-hour volume of 48.0 billion, and bitcoin dominance sits at 58.84 percent with ether at 11.57 (CoinGecko, live Sunday).

CoinGecko's trending list reads LSK, EMBER, PONS, STONK, PENGU, ZEC, LAPTOP, SOL, PI and ZCAT. Zcash is the one with a price attached to it: it trades at 1,079 dollars on Hyperliquid with 512.3 million dollars of open interest, the fifth largest book on that venue, and it fell 5.98 percent on the day. CoinDesk headlined Friday that zcash led losses as traders bet on a hike, then that bitcoin recovered as zcash leverage unwound. Ninety-five of the CoinDesk 100 declined on Friday.

Derivatives

Funding is the cleanest disagreement on the board. Hyperliquid pays shorts on every major: bitcoin at minus 12.21 percent annualised, ether minus 4.28, Solana minus 12.90, XRP minus 22.28. Binance disagrees on bitcoin and ether only: bitcoin pays plus 0.0026 percent per eight hours, or plus 2.85 percent annualised, and ether plus 3.53. Solana is negative on both venues, minus 4.76 percent on Binance against minus 12.90 on Hyperliquid, so the sign flip is a two-asset story rather than a three. Total Hyperliquid open interest is 10.044 billion dollars across 234 markets. A caution for anyone reading the Hyperliquid alt book: a long tail of names prints funding at exactly 10.95 percent annualised, which is that venue's structural base rate and not a positioning signal.

Aggregate futures open interest across venues fell to 59.5 billion dollars from 62.4 billion. Liquidations ran 256.3 million dollars in 24 hours against 142.3 million midweek, and more than 140,000 traders were liquidated during Thursday's drop. Friday's ether squeeze carried between 427 and 668 million dollars of short liquidations depending on the aggregator, and this desk prints the range.

Options

The 25 September expiry is the whole board: 185,048 bitcoin contracts, 44.8 percent of all open interest, with a desk-computed max pain of 72,000 dollars, which is 6.17 percent below spot. Ether's same expiry carries 741,039 contracts, 42.4 percent of its board, max pain 2,200, which is 11.36 percent below spot. The nearer 18 September expiry, which lands on the Bank of Japan decision, carries 17,271 bitcoin contracts at a 78,000 pin and 95,129 ether contracts at 2,500.

A caveat this desk owes on max pain. The mechanism depends on dealer hedging into expiry, and that is strongest in the last day or two before settlement and weak twelve days out. A good deal of Deribit open interest also sits in directional and structured positions rather than hedged dealer books. Treat 72,000 as a gravity reading, not as a level with the standing of support or resistance.

Max pain is computed here from Deribit's full per-instrument open interest by minimising total in-the-money value across every listed strike, not taken from a vendor. It reconciles exactly with the 72,000 pin this desk published on 10 September, which is the check that the method is stable.

Traditional Markets

American equities rose on Friday and fell on the week, which is the whole mood in one line. The S and P 500 closed 7,656.98, up 0.86 percent on the session and down 0.80 percent on the week. The Dow added 0.98 percent to 52,573.29 and lost 1.57 on the week. The Russell 2000 gained 0.45 percent to 2,903.94 and lost 2.41 percent on the week, the worst of the majors, which is the only part of the equity tape behaving as though rate policy matters (CNBC quote API at the Friday stamp, prior-week closes from CBOE daily files).

The curve

The move is at the front. Across Wednesday to Friday the two-year added 20 basis points, the three-year 20, the five-year 17, the ten-year 13 and the thirty-year seven. On Friday alone the thirty-year fell two basis points while the two-year rose seven.

Spreads on Friday: two-year to ten-year plus 33 basis points, from plus 40 on Wednesday. Two-year to thirty-year plus 72, from plus 85. Twenty-year to thirty-year is inverted at minus three. Three-month to ten-year plus 89. The two-year at 4.63 percent sits a full 100 basis points above the 3.63 percent effective funds rate (US Treasury daily yield curve; effective rate and the operative 3.50 to 3.75 percent target range pinned at the New York Fed reference-rates API, which publishes the range alongside each daily print).

The Fed, with the arithmetic shown

Current target range 3.50 to 3.75 percent, effective rate 3.63 on 108 billion dollars of volume, SOFR 3.62 on 2,921 billion. The last move was a cut on 11 December 2025 and there have been no 2026 changes. The meeting is Tuesday and Wednesday 15 to 16 September with a Summary of Economic Projections attached.

Probability-weighted post-meeting midpoint is 3.8388 percent against a current 3.625, so 21.4 basis points are priced for Wednesday. October buckets run 7.2 percent hold, 49.6 percent at 3.75 to 4.00 and 43.2 percent at 4.00 to 4.25, a weighted midpoint of 3.9650 and 34.0 basis points above today. The market is not pricing one and done. Classification here is done by comparing each bucket floor with the current target floor rather than trusting a vendor's hold-cut-hike labels, which have been mis-rendered before.

Commodities and currencies

Crude is the week's event and the daily print hides it. West Texas ran 9.37 percent and Brent 8.65 percent across five sessions, including a single 6.69 percent session on Thursday, before giving back 2.37 and 2.81 on Friday. Brent's premium to West Texas narrowed to 4.56 dollars from 5.15 on Thursday. One correction is owed here, and it is a real one. This desk published a 108.19 dollar Thursday Brent settle on 10 September, with a 6.90 percent daily gain and a 10.49 percent two-session move. All three figures were wrong, and they were a single error rather than three. The Wednesday settle of 101.21 multiplied by 1.0690 gives exactly 108.19, so an intraday percentage was applied to the previous day's settle instead of Thursday's settle being taken directly. Brent settled at 107.63 on Thursday, up 6.34 percent, and the two-session move from Tuesday's 97.92 was 9.92 percent, not 10.49. Friday settled at 104.61, down 2.81 percent, on ICE Brent November 2026. World Oil published 103.51, which sits one cent above Friday's session low of 103.50 and is a late-session intraday print rather than a settlement, so it is discarded rather than averaged. The weekly gain from 96.28 on 4 September is 8.65 percent.

In currencies the only real move is the yen. Dollar-yen fell 1.73 percent on the week to 153.54, and Reuters reported Sunday that speculators have turned net long yen for the first time since February. The dollar index is unchanged on the week at 99.095. The Swiss franc weakened 0.83 percent.

International Markets

This issue has been saying four central banks in nine days, and that count is too small. It is right for the window itself, but Asia moved first and moved earlier. South Korea has already hiked twice, to 3.00 percent on 27 August. Singapore tightened its currency band on 27 July with more expected in October. Counting every authority that has repriced or will within five weeks, at least six have moved, and the first movers were neither American nor European. A brief that opens on the Federal Reserve is describing the loudest part of this cycle rather than the start of it.

The sequencing matters and almost every write-up will get it wrong. Every Asian market closed before the American consumer price index was released. That release came on Friday morning New York time, by which point Tokyo, Hong Kong and Mumbai had all already closed for the week. So Friday's Asian selloff was a reaction to Thursday's producer prices, the oil move and the bond rout, and contains no information about the consumer price print at all. Monday is Asia's first chance to trade it, and it arrives alongside the weekend escalation in the Gulf.

The Swiss market is the worst major on the week at minus 4.31 percent, with a zero percent policy rate and a franc that weakened 0.83 percent. An earlier draft called that an unusual combination and left it there. It is not unusual, it is the yen mechanism running in reverse. Switzerland is the one major economy not tightening at all, so as every neighbour raises rates the franc becomes the region's cheapest funding currency, and a currency being borrowed to buy something else weakens even when it is the safe one. Korea is the only large gainer at plus 3.33 percent, carried by a single 4.61 percent session on 7 September.

India deserves more than a table row, because it is the most on-thesis emerging-market story in the issue. The rupee lost 1.13 percent, the worst in the region, and India imports most of its crude. That is the Hormuz shock arriving in a currency a full cycle before Monday's August inflation print confirms it, which a Reuters poll expects at a twenty-month high. Note what this brief does not print: an August figure. A widely used calendar shows an August actual of 4.45 percent beside a release date of Monday, which is tomorrow. That is the July number in a row offset by one month, and the same vendor has been caught publishing future-dated actuals before.

Geopolitics

The direction is escalation, on two fronts, and one correction to how this desk has been framing it. Hormuz is not a risk-of-closure story. Traffic has been reduced to a trickle since roughly 20 June, with seven vessels transiting on Thursday against a ten-day average near fifteen, which this desk treats as closure in practical rather than absolute terms, and says so because the transit count is not zero. Iran closed it and has set terms for reopening. This is also a United States and Iran war, which means the Israel and Iran ceasefire contract resolving 30 September is a separate instrument from the conflict driving the barrel.

Friday: the bypass went down

Saudi Arabia shut the East-West pipeline, the line whose entire purpose is to move crude to the Red Sea without transiting Hormuz. Drones struck it on Thursday and the shutdown was announced Friday evening (Financial Times, Bloomberg and Reuters, all on the evening of 11 September). Maritime Executive confirmed substantial impacts at two pumping stations. Capacity is reported at 7 million barrels a day by Reuters and Al Arabiya earlier this year, and at up to 5 million by Business Insider Africa on 12 September. It is the second strike on this line in 2026; Iran hit it on 8 April. The same day the Houthis reached Perim island at the mouth of the Bab al-Mandab, halving traffic there to about 26 vessels a day, and Hormuz transits ran seven vessels against a ten-day average near fifteen.

Saturday: attribution and a political clock

Iraq confirmed the pipeline drones launched from its territory and dismissed two senior officials (Al Jazeera and Associated Press). Trump said Iran was probably responsible, and separately that the war will end immediately after the United States midterm elections, predicting oil falls sharply then (CNBC). The Houthis formally seized Mayun Island. Bahrain declined to attend Iran's proposed Hormuz meeting. Iran handed the United States seven conditions for reopening the strait; this desk could not retrieve the conditions themselves from any fetchable source and records that as a gap rather than guessing at them. Iranian state media stated plainly that the Iran and Oman understanding does not mean reopening. The BRICS summit in New Delhi, with Putin, Xi, Modi and Pezeshkian present, urged maximum restraint and condemned attacks on civilian infrastructure.

Sunday: a ship, hours before the talks

An Iranian commercial vessel was struck off Qeshm Island near the strait. One killed, three wounded on most outlets, four on Iranian state media, and this desk prints the conflict rather than averaging it. Iran attributes the strike to the United States. There was no American or Central Command confirmation or denial as of Sunday midday, and the pattern is worth stating: the United States denied Iranian strike claims on both 6 and 9 September. The vessel was hit hours before Monday's Hormuz talks in Oman. Separately Iran and Oman agreed the geometry of new entry and exit routes whose entry point lies entirely in Iranian waters, which is a routing arrangement and not a reopening. Reuters reported no signed deal is expected Monday, while Anadolu reported Arab foreign ministers would sign one, and those two reports, each apparently sourced to Iranian officials, cannot both be right.

A Houthi projectile wounded two and damaged a mosque in Saudi Arabia's Jazan region. Iraq seized a drone-launching platform used against the pipeline. Roughly 1,400 Yemenis fled to Djibouti inside 24 hours.

One transmission channel this desk had been ignoring. The barrel is not the only price rising. Tanker rates are at record highs, with Gulf very large crude carriers referenced near 650,000 dollars a day, and war-risk premiums on these routes rise with every seizure. Those costs reach the consumer as freight rather than as crude, which means they arrive in goods prices on a longer lag and are invisible in the oil price the bond market is watching. A brief that reads inflation off Brent alone is reading one channel of two.

Regulatory

The CLARITY Act faces a cloture vote Tuesday and the arithmetic says it fails

The Senate takes a procedural vote on the CLARITY Act on Tuesday 15 September, confirmed independently by six outlets including CoinDesk, Blockhead and the Bitcoin Foundation, several citing the Senate schedule. Both congress.gov and senate.gov returned HTTP 403 to this desk, so the date is press-confirmed rather than clerk-confirmed, and confidence is 9 out of 10.

The math is not close. Cloture needs 60 votes. Republicans hold 53 seats, so at least seven Democratic or independent crossovers are required. As of 11 September zero Democratic senators support the revised text, a 630-page rewrite released days before the vote. Only Senators Ruben Gallego and Angela Alsobrooks backed the bill at committee stage, which is two, not seven. Treasury Secretary Bessent publicly urged the Senate to advance it. Speaker Johnson cut two weeks from the House schedule on 3 September, so even a surviving bill faces scarce floor time.

Polymarket prices CLARITY signed into law in 2026 at 22.5 percent on 15.30 million dollars of volume, which calibrates to 16.5. The market and the whip count agree, and this desk agrees with both.

Enforcement

The CFTC rule is the one with a second-order effect. A 30 percent presumptive award on recoveries of five million dollars or less makes small-case whistleblowing economically rational for the first time, and the incentive lands on exactly the mid-size venues that previously sat below the enforcement threshold. Expect more referrals, not fewer.

On trade and export controls there was no new action inside the window. That is worth stating plainly rather than recycling standing tariff policy as though it were news.

Scandals and Crisis Events

The window was not quiet. Eight exploits, one breach that matters more than its size, and a heavy enforcement calendar.

Blockstream Liquid remains the largest crypto hack of 2026: roughly 4,000 bitcoin, about 320 million dollars, exploited 6 September, with approximately 600 bitcoin still outstanding. Block production resumed on 10 September but peg-outs remain frozen, and on 11 September Blockstream publicly refused the ransom demand, saying plainly that it is theft (Decrypt, The Block, Cointelegraph; Chainalysis published a mechanism post on 9 September). A group calling itself Bitcoin Red Team claims Blockstream ignored prior warnings. The exposure is federated-peg trust across Bitcoin sidechains rather than any single token.

Revolut disclosed a breach on 12 September in which an unauthorised third party used a legitimate government agency email domain to submit forged data requests. Taken: dates of birth, addresses, passports, driving licences, possibly verification selfies and account statements, and bitcoin transaction records. Revolut called the number of affected customers limited and refused to give a figure to TechCrunch. Neither the agency nor an occurrence date has been disclosed. The consequence is deanonymisation: it ties on-chain activity to home addresses for an unknown number of retail holders.

A sourcing note that kept two bad items out of this brief. The collector found that CoinMarketCap and Yellow.com pushed at least five months-old or years-old events into news feeds with fresh publication dates this window. The Nesa exploit is the clearest case: it was published on 11 September and actually occurred on 24 August, confirmed from inside the report body. Event dates here come from article bodies, never from feed timestamps. This desk published a retraction on 9 September that rested on exactly this failure, and the rule now applies to retractions as strictly as to claims.

On-Chain and Whale Intelligence

The fund flows are the cleanest signal in the digital-asset complex this week, and they point in opposite directions for the two largest assets.

Monday 7 September was Labor Day, so the week holds four sessions, not five. Bitcoin funds printed an outflow on every one of them and ended a three-week inflow streak. Holdings stand at 1,261,711.89 bitcoin, 6.282 percent of market capitalisation, against total net assets of 97.58 billion dollars. Only three of twelve funds reported any flow on Friday, and BlackRock's IBIT was the largest single outflow at 19.23 million.

Ether's Friday was the largest single day in the visible series at 216.41 million dollars, of which BlackRock's ETHA took 148.82 million. Holdings are 6,442,746.72 ether, 5.279 percent of market capitalisation. Every ether fund that reported a nonzero flow reported a positive one. Ether fund value traded on Friday was 2.56 billion dollars, 3.3 times the Thursday figure and roughly equal to bitcoin's for the first time in the series.

One structural detail, reported with its competing explanation rather than without. Eight of eleven ether funds closed at a discount to net asset value while eleven of twelve bitcoin funds closed at a premium. The tempting reading is that money is arriving in ether through vehicles the market still marks below fair value. The duller reading is at least as likely: discounts are common in newer and smaller products with thinner authorised-participant activity, and because digital assets trade continuously while net asset value strikes at a fixed hour, a weekend round trip of the size both assets just made will produce exactly this pattern mechanically. This desk cannot separate the two and is not going to pretend the flattering one is established.

Data integrity note. The provider's two endpoints disagree on cumulative bitcoin inflow by 21.4 million dollars, 55.176 billion against 55.155 billion, while assets under management agree exactly. There is no equivalent discrepancy in the ether series, which means this is specific to the bitcoin series rather than a systematic mark-to-market versus flow-only offset, as this desk previously assumed. Both figures are printed and neither is chosen.

Stablecoins, chains and network

Total stablecoin float is 311.60 billion dollars. Tether is 183.49 billion, up 0.09 percent on the week. USDC is 74.35 billion, down 0.48. The growth name is Ethena's USDe at 4.60 billion, up 5.76 percent on the week and 16.27 percent on the month. That deserves a flag, and an earlier draft left it as an unanswerable one. It is answerable. USDe's yield comes from perpetual funding, and funding on Hyperliquid has gone negative across every major. But Ethena hedges across venues, and the venues carrying most of its notional are still funding-positive on bitcoin and ether, so one venue turning negative does not invert a blended carry. There is a staking and Treasury-bill sleeve underneath it as well. The likelier driver of 16 percent monthly growth is leverage looping and points farming in lending markets rather than fresh capital chasing a carry that has narrowed.

Chain total value locked: Ethereum 50.07 billion, Solana 5.90, BNB Chain 5.62, Base 5.58, Tron 5.48, Bitcoin 4.17. Bitcoin network hashrate is 943 exahash with the next difficulty adjustment estimated at plus 4.72 percent, and fees are at the floor of 1 satoshi per vbyte, which is its own comment on on-chain demand.

Sovereign and Institutional Flows

The auction tape is the most under-read data of the week, and it contradicts the story the yield level implies.

Every percentage above is computed from the raw dollar awards rather than taken from a vendor, and all 61 auctions in the 60-day window sum to exactly 100.00 percent as a validation check. Dealer takedown of 2.21 percent on the thirty-year is the lowest of any coupon auction in the window by a wide margin; the next lowest is 8.60 percent. Dealers absorb what end users decline, so a 2.21 percent residual means end users did not decline. Both September long-end reopenings cleared above their August equivalents in yield while taking far more indirect demand, and the Federal Reserve's own add-on was zero at every September auction after running between 2.0 and 14.7 billion dollars through July and August. The demand was private and it was real.

This is the single best argument against reading a 5.35 percent thirty-year as a fiscal-credibility event. The long end is repricing policy, not credit.

What this desk could not source: China's latest official gold tonnage at filing grade, Treasury International Capital holdings with a month-over-month delta, and any sovereign wealth fund disclosure inside the window. Those are gaps, not zeros, and they are named here rather than filled with press coverage.

Prediction Markets

Disclosure first. On 11 September the Commodity Futures Trading Commission authorised three previously unreported insider-trading investigations into Polymarket, covering its Biden pardons, Iran and Google contracts markets, originally reported by WIRED. This brief quotes Polymarket prices throughout, including in its lede. No traders have been named and these are investigations, not charges. Readers should discount this venue's geopolitical markets accordingly, and this desk will do the same.

The calibration earned its place this week. Polymarket's raw 78.5 percent on a September hike calibrates to 84.5, and CME-derived futures buckets independently print 85.5. A seven-point gap between two venues closes to one point after correction, which is what the adjustment was fitted to do and is not usually this clean.

The two most informative prices are not about the Fed, and one of them has to carry the discount declared above. Hormuz traffic returning to normal by year end trades at 19.5 percent, which would mean the market expects the disruption to persist through December, a very different assumption from the one in most 2027 earnings models. That is precisely a geopolitical market on this venue, so it is reported here as an indication and explicitly not used as evidence anywhere else in this brief. CLARITY at 22.5 percent is the price this desk does lean on, because it is corroborated independently by a Senate whip count rather than standing alone.

Kalshi is a stated gap. Its Fed series enumerates 60 markets across every resolved address but returns null on every quote field. That is not the Swiss resolver problem documented previously; the addresses resolve and the market list returns. Pricing appears to require authentication. This desk reports it as a gap rather than claiming the venue is dark.

Sentiment and Positioning

The single most useful line in that table is the last one nobody watches. SKEW rose 7.47 points to 154.49 on Friday and was the only volatility series that went up. At-the-money volatility was sold hard while tail protection was bid. That is not complacency, it is something more specific: desks are confident about the central case and are paying up for the left tail. A market that sells the body and buys the wings is telling you it expects the Fed to do exactly what is priced, and is quietly hedging what happens if the guidance is worse than the decision.

One honest qualification, because this indicator is routinely oversold. SKEW is noisy and its standalone predictive record is poor; it is cited here as one reading consistent with the option-adjusted spread and nine-day volatility evidence, not as a decisive tell on its own. A single Friday print is a straw in the wind, and the wind is what the other two series are describing.

The tail risk that follows from it, and which this brief would otherwise have missed: if at-the-money volatility has been sold while tails were bought, the dealer book going into Wednesday is short gamma near spot. A short-gamma book amplifies a surprise rather than damping it, so the same positioning that makes the central case look cheap also makes the miss more violent than the odds alone suggest.

Cross-asset linkages

They are the same shape and not the same event. Germany's flattening is confirmation after a delivered hike, by a council whose own staff projection has core inflation still rising into 2027. The American flattening is anticipation of a hike that has not happened and can still be withdrawn. Identical curves, opposite epistemic status, and only one of them is reversible on Wednesday afternoon.

Germany bear-flattened on Friday exactly as the United States did, 8 basis points at 2s10s and 13 at 2s30s, with the two-year up 13 and the thirty-year unchanged. Two continents, same shape, same week. The United Kingdom carries a 30-year to 40-year inversion of about 30 basis points that has persisted all week, the same structural family as the American 20s30s. Japan is the only major curve with no inversion anywhere, and it is the one whose central bank has the furthest left to travel.

Cross-market ten-year spreads: the United States over Germany is 140 basis points, Germany over Japan 64, the United Kingdom over Germany 185. Euro-area periphery stress is widening slowly and monotonically, with the ten-year all-government premium over AAA moving from 42.3 to 44.9 basis points across four sessions. Small, but it only went one way.

The credit answer, stated plainly. Exchange-traded proxies ranked the week's damage as long Treasuries worse than investment grade worse than high yield, down 1.63, 1.10 and 0.71 percent respectively. That is the duration ordering, not the credit ordering, and actual option-adjusted spreads moved between 1 and 4 basis points all week. This is a rates event, not a credit event, and the tightening in spreads is a real signal rather than a stale mark. No 11 September spread print exists on any free route because the index publishes a day late, and that limitation is stated rather than papered over.

Sovereign credit default swaps, a route recovered

This gap was closed during production. Five-year sovereign protection, in basis points: Egypt 266.5, Turkey 221.5, South Africa 116.9, Brazil 111.3, Indonesia 82.6, Mexico 80.1, Saudi Arabia 57.2, Israel 53.7, France 37.5, China 34.8, Italy 34.3, United States 32.0, Japan 23.1, South Korea 21.6, United Kingdom 18.7, Spain 16.1, Germany 7.2, Switzerland 7.0. Quote dates run from 9 to 11 September and are not uniform, which is stated because it matters for the comparison.

The line worth stopping on is France at 37.5 trading wider than Italy at 34.3. The core and periphery order is inverted, and France also widened 2.5 basis points while Brazil and Mexico tightened. Saudi Arabia at 57.2 is the other one to watch, because it is the sovereign whose pipeline was struck this week and whose output is at a 1990 low, and its protection has barely moved. Either the market has not repriced Saudi risk or it believes the disruption is genuinely temporary. Note this is a single vendor with no second source, so it is reported as indicative rather than as a confirmed level.

Still missing: foreign-exchange implied volatility has no free route this desk can reach; the New York Federal Reserve discontinued its page in 2013. That is where a compressed-volatility regime would show strain first, and its absence remains a real hole in the analysis above rather than a rounding error.

The Bottom Line

Last week's views, marked

The 10 September issue carried six trade views, four of which its own bias auditor found were expressions of a single factor. The Brent card is ACTIVE and ahead, having been carried by an 8.65 percent weekly move in the barrel even though Friday closed lower. The gold card is ACTIVE and behind, with gold down 1.51 percent on the week at a 4,408.90 settle. This desk does not reproduce the prior entry, target and stop levels here because it could not verify them from the published issue, and states that rather than inventing them. The factor-concentration warning from that issue is repeated above because it has not been fixed, only disclosed.

Trade views

On concentration, and the panel did not agree with itself. An earlier draft claimed five of six views rest on the same policy-rate factor. One reviewer put the true figure at three, on the grounds that Brent is driven by physical supply, CLARITY by a Senate whip count and the ether view by fund flows. Another put it back at five, arguing Brent transmits through the dollar and risk appetite and so is not independent on a Federal Reserve day. This desk cannot settle that and prints both: on the narrow count three views die together if Wednesday surprises, on the wide count five do, and only the CLARITY view is independent on either reading. The risk reviewer sized the wide case at 60 to 70 percent of the book's risk budget lost in one session on a hold, which is currently priced at 14.5 percent.

Where the discount applies it is marked on the card itself rather than asserted here, because a conviction number that moves only in the prose is exactly the failure a previous issue was correctly criticised for.

Catalyst calendar

Review Panel

Eleven reviewers audited this issue, and their findings arrived in two waves, covering the fourteen standing lenses with three pairs merged: regulatory with geopolitical, international with sovereign, and portfolio with sentiment. The panel changed the brief substantially, and the largest change was to its central claim.

The lede was rewritten, not softened. The draft claimed that a rising real yield against a falling breakeven settled the question against the energy explanation. The Devil's Advocate and the Bias Auditor independently established that it does not: a supply shock large enough to destroy demand produces the identical signature to credible tightening, because both raise real yields and lower inflation compensation. The draft picked the comfortable reading. The lede, the second story and the second Bottom Line theme now state that the print rules out simple energy pass-through and rules out nothing else, and the issue is organised around the fork rather than around a resolved contradiction.

An asset-mixing error was caught in the positioning story. The draft paired bitcoin's retail long-to-short ratio with ether and Solana's top-trader ratio and called the result a retail-versus-professional divergence. Checked within each asset, bitcoin shows no divergence at all; its top-trader ratio rose alongside retail, from 2.1312 to 2.2324. The genuine divergence is in ether and Solana, which relocates the cascade risk to ether, where the fund inflows are arriving, and away from bitcoin, where the draft had placed it.

The auction claim was downgraded from proof to evidence. The Flow Analyst established three mechanical objections: both September prints were reopenings, which attract index and basis demand regardless of sentiment; indirect is a bidding channel rather than a buyer type and can contain repo-financed leveraged demand; and Treasury discloses no bidder identity, so a 79.48 percent award could be one account. The same reviewer noted the brief applied single-allocator caution to a 216 million dollar ether print and withheld it here. The claim now rests on the independent spread data, with the auction as corroboration.

The concentration discount was found to be cosmetic, and the panel then disagreed about the underlying number. The Bias Auditor showed the one-pip haircut existed only in prose and never on a card, and that the claimed five of six was itself inflated. The Risk Manager put it back at five on transmission grounds and sized a Federal Reserve hold at 60 to 70 percent of the book's risk budget. Both readings are now printed, the discount is marked on the individual cards, and two views were reclassified out of the trade list entirely: the ether view because its own falsification test has not run, and the 20s30s segment because a view that cannot be stopped is an observation.

Four arithmetic errors were found and fixed, all in the prediction-markets table and the flow prose: two calibration edges rounded the wrong way, one calibrated probability was understated, and the weekly bitcoin fund total disagreed between table and text. On the last of these the reviewer prescribed the wrong direction; the exact sum is 462.733 million dollars, so the table was right and the prose was corrected to match. Every date and weekday in the issue was recomputed and all eighteen were correct.

Two gaps were closed during production and one remains. A sovereign credit default swap route was recovered and appears in the cross-asset section, where it shows France trading wider than Italy. A freight and war-risk insurance channel was added to Geopolitics as a second inflation transmission route the draft had ignored entirely. Foreign-exchange implied volatility still has no free route and is stated as a hole rather than filled.

Second round, after first publication

A first version of this issue was published before four reviewer reports had been read in full. Reading them produced a further round of corrections, three of which were factual errors contradicted by the brief's own tables, and this version replaces the first.

Funding. The draft said Binance funding was positive on the same assets where Hyperliquid was negative. Its own derivatives table shows Binance Solana funding at minus 4.76 percent. The cross-venue sign flip covers bitcoin and ether only, and that is now stated in both places the claim appeared. Attribution. The draft described Reuters and Anadolu as Iranian-sourced accounts. They are not Iranian outlets; each was reporting a claim sourced to Iranian officials, and the sentence now says so. Hormuz. The draft asserted flatly that the strait is closed, two sentences before printing a transit count of seven vessels. It now reads as closure in practical rather than absolute terms, with the count stated.

The lede was refined a second time. The first revision replaced a false certainty with a two-way fork, credible tightening against demand destruction. The macro reviewer showed that demand destruction normally pulls real yields down, so that fork was also wrong. There is a third and better-fitting explanation: the price data that triggered the repricing was itself partly energy, which makes policy and energy one story arriving through the central bank rather than two stories competing. The issue now says that.

The economist withdrew the Keynesian label as generic cost of carry, the same failure as the Austrian label withdrawn last issue, and supplied the supply-shock framing that replaces it. The international desk showed the four-central-bank count contradicted this brief's own table, since Korea and Singapore moved first, and solved the Swiss anomaly the draft had named and abandoned. The crypto reviewer supplied the derivation of the 10.95 percent funding floor, the caveat that max pain is weak twelve days from expiry, and the answer to the Ethena question the draft had left open. The unverified forecaster survey no longer carries any argument.

Two corrections against this desk's 10 September issue are printed in the sections above: the Brent Thursday settle, where the mechanism of the error has been identified precisely, and the United Kingdom thirty-year gilt claim, which conflated an auction clearing yield with a secondary market level.