The Treasury bought the long end one session. Crude is taking it back before the bell.

The fix held for one close

1. A third of the long-end rally is already gone
   The thirty-year fell to 5.194% on the buyback announcement and is higher again this morning; the ten-year has given back roughly half its move. Nothing about the announcement changed overnight. Section 07.
2. Crude is the overnight story, not the buyback
   WTI +2.92% and Brent +2.56% after the White House announced what it called an unprecedented economic operation against Iran. Hormuz has been effectively shut since February. That is an inflation input the buyback cannot reach. Section 07.
3. The two-year is trading above the funds rate
   The two-year sits some forty-five basis points above the top of the policy range. The front end is priced for tightening, and the July minutes carried three dissents, all to hike. Section 09.
4. Rieder took the other side of the front end, on the record
   BlackRock’s CIO of global fixed income told CNBC the first move Chair Warsh makes is a cut, not a hike — and said where he is positioned to express it. That is the sharpest disagreement on the tape. Section 05.
5. Wednesday was a debasement rally wearing an equity costume
   Metals, crypto and the grains all settled up multiple percent against an S&P up 0.21% and a Nasdaq 100 that fell. The index was close to the worst thing you could have owned. Section 10.
6. Claims and Philly Fed land together at 8:30
   Two second-tier prints become first-tier when the front end is pricing a hike and the last Philadelphia Fed reading came in at −0.3 against a consensus near +6.8. Section 03.

Where everything sits

Bitcoin trades around the clock and its twenty-four-hour change depends where you start the clock, so the settled Wednesday session move is the figure shown.

What the gauges say

The flow read

Positioning is the uncomfortable part of this tape. Trend-following exposure to equities sits near the 94th percentile of its own history, with roughly $90 billion of global equity length on Goldman’s estimate after last week’s modest de-risking was almost entirely reversed. That is not a contrarian signal on its own — systematic length can stay extreme for months — but it is why a vol shock and a selloff arriving together behave differently from either arriving alone. Against it, the corporate bid is returning: most S&P reporters cleared their earnings blackout by the second week of August, and 2026 buyback authorisations are running at a record annualised pace. Wednesday’s tape — index up a fifth of a percent, metals up four — is what that combination looks like when the money has somewhere better to go.

Calls graded

Two second-tier prints in a first-tier position

No coupon supply today and no Fed speaker to interpret anything. What there is instead is a pair of releases that would normally pass without comment, landing together at 8:30 into a front end priced for a rate rise. Initial jobless claims have been drifting up — the last print came in at 209,000 against a 202,000 consensus — and the Philadelphia Fed manufacturing index delivered one of the uglier misses of the summer at its last reading, −0.3 against a consensus near +6.8 and a July print of +15.9. The Conference Board leading index follows at 10:00 and EIA natural gas storage at 10:30, into a contract already down two percent this morning.

The reason a claims number matters more than usual is the asymmetry in Section 09: this committee’s last recorded action was three people voting to tighten, and the market prices roughly a one-in-three chance they get their way in September. Labour data is the only input with the standing to move that. Neither print decides the September meeting, but together they are the first evidence since the announcement that could tell you whether the long end is rising on growth or on supply — which is the whole argument in Section 07.

Scenario language describes what has typically followed similar configurations. It is not a recommendation and not a forecast.

The rest of the week

Walmart reported before the open this morning and Ross Stores reports after the close, closing out a retail week that has already split down the middle — Section 08. Friday carries no top-tier US release. The event the whole tape is actually leaning toward is next week: the Kansas City Fed’s Jackson Hole symposium runs August 27–29, themed on financial innovation and payments, with Chair Kevin Warsh delivering his first keynote as chair on the Friday morning. That is five sessions from now, not one — worth stating plainly, because a good deal of this week’s commentary has been written as though it were imminent.

The flip fell to meet a rising tape, and landed on support

Two things moved on Wednesday’s close and they moved toward each other. The dealer gamma flip fell from 7,703.01 to 7,678.98, a twenty-four-point drop, while cash rose sixteen points to 7,707.98. Yesterday the index sat eleven points below the flip — the first negative-gamma reading this letter has logged. This morning it sits twenty-nine points above it, and the regime is positive again. Two crossings in two sessions is the honest description of where structure stands: this is not a market with a comfortable cushion, it is a market whose regime line is close enough to change hands on an ordinary day.

And the corridor tightened again. The call wall came in from 7,800 to 7,750 and the put wall rose from 7,600 to 7,640 — a band that was five hundred points wide a week ago, two hundred yesterday, and is a hundred and ten this morning, with cash sitting nearer the top of it than the bottom. Neither boundary is far enough away to be ignored on an ordinary session. The heaviest call open interest on the board is higher still, at 7,800, but the net gamma there is thinner because put open interest at that strike is nearly as large — which is why the wall sits where it does rather than where the raw call interest is.

Put the two together and the shape of the day is legible. The ceiling sits forty-two points above cash. The flip is twenty-nine below, and the put wall a further thirty-nine under that. Cash is therefore in the upper half of the band, nearer the level that caps it than the level that would accelerate it — the configuration that tends to produce a narrow, grinding session. The risk in it is on the other side: there is no structural level at all between the flip and the put wall, so if the flip goes there are thirty-nine points with nothing in them.

Dealer gamma map — SPX

Strike-level detail runs from 7,610 to 7,805. Both walls are local peaks with live strikes on either side of them, but a heavier strike outside that band would not be visible from here.

The number that matters most is a coincidence of two independent sources. Converted at the 21.02 basis, the flip lands at 7,700.00 in the September contract. Cannon’s first support, built on an entirely different method, is 7,695.08. Two unrelated frameworks land five points apart, which makes 7,695–7,700 a single band rather than two lines: the flip gives way first and support second, and underneath both, dealers stop damping and start amplifying. The next Cannon support is not until 7,663.42 — thirty-two points of comparatively empty space.

Above, the picture is quieter. ES opens on the pivot at 7,729.92, first resistance 7,761.58, second 7,796.42, with the August high at 7,838.50 set on the 13th. Newton’s new-highs call needs that level inside five sessions, through both resistances and into the heaviest call strike on the board.

Cannon Daily Levels — August 20, 2026 · September contracts unless noted

Cannon Edge — Wednesday’s settlements · short- and long-term trend reads

The Edge trend column is worth a second look this morning. The thirty-year bond carries a down read on both the short- and long-term trend — on the morning after the single largest official intervention in that market this year, and after a session in which the contract gained 1.35%. Natural gas carries the same double-down. Everything else on the board that moved on Wednesday — gold, silver, copper, the grains — reads up on both. The trend framework has not yet accepted that Wednesday changed the long end’s direction, which is the same conclusion the overnight tape has reached by a different route.

New and moved voices

The full roster

Sorted by how much your day changes if the voice is right. Names who said nothing this week are listed anyway, because on a day like this silence is a position too.

You can buy the bond. You cannot buy the barrel.

Start with what the Treasury actually did, because the headline has been loose. The maximum size of liquidity-support buybacks for longer-dated nominal coupon securities roughly doubles, from a prior cap of $2 billion to at least $4 billion per operation. It covers the ten-to-twenty and twenty-to-thirty year sectors, and the number of long-end operations rises from two per quarter to four. It takes effect on September 9 and runs through November 4, the end of the current refunding quarter. Those are per-operation caps on repurchases. They are not a target, not a forecast, and not a commitment to a yield level.

The market treated it as one anyway, and that gap between what was announced and what was priced is this morning’s problem. The thirty-year fell nine to ten basis points to 5.194%, having printed 5.34% the previous session, a nineteen-year high; the ten-year fell about six; the dollar broke down. By this morning the thirty-year has retraced roughly a third of that and the ten-year about half, with no new information. A duration operation of a few billion a fortnight, beginning in three weeks, is a real bid. It is not large against net issuance, which is El-Erian’s point, and it does not change the quantity of paper that has to clear.

Which brings in the barrel. WTI is up 2.92% this morning and Brent 2.56%, and this is not a positioning move. The Strait of Hormuz has been effectively closed since fighting began in late February, with transits running at a fraction of the roughly 130 a day that used to pass. On Wednesday the White House announced what it called an unprecedented economic operation against Iran, threatening secondary consequences for any country still doing business with Tehran; Iran’s position is that the Strait stays shut until sanctions ease. Whatever the diplomacy does next, the price is the transmission channel, and it runs into the inflation print the front end is already pricing a hike against.

That is the pincer. The Treasury can lean on the long end mechanically. It has no instrument that touches crude, and crude is the input most likely to keep the three dissenters in the room. A curve where the long end is managed downward while the short end is pushed up by an oil-driven inflation impulse is a flattening curve for the wrong reason — though note 2s10s did the opposite this morning, steepening, because the long end gave back more than the front end did.

The dollar is the release valve. Saravelos put it precisely in Section 05: if the bond price is not allowed to adjust, the currency adjusts instead. The index is down 1.23% in five sessions and 2.20% over the month, and every part of Wednesday’s leadership — gold, silver, emerging-market currencies, crypto, the grains — is a variation on that one trade. Natural gas is the exception, down 2.03% this morning with a down read on both Cannon trend horizons, which is a useful reminder that this is a currency story rather than an energy story.

And the supply behind it keeps growing. Federal debt has passed $40 trillion. El-Erian’s framing on air was that the country now carries a massive lever back on future growth and productivity, and separately that hyperscaler issuance has no visible limit — it must continue, must be absorbed, and comes at the cost of high borrowing costs. The buyback is a bid for a few billion at a time against that. It is not nothing. It is not the same order of magnitude as the thing it is standing in front of.

The single names, told once

The consumer split cleanly on Wednesday and both halves were real. Target beat and raised — comparable sales up 3.8% against a 2.4% consensus, digital up 8.7%, full-year guidance lifted — though the headline earnings figure carries a substantial tariff-refund benefit that flatters the comparison. Lowe’s went the other way, trimming full-year sales to the bottom of its range and cutting the comparable-sales outlook to flat, citing weak discretionary do-it-yourself spending. Home Depot had beaten on Tuesday. El-Erian read the split as the low-income household under pressure rather than as a sector story, which is the right frame: the bifurcation is in the customer, not the retailer.

The healthcare shock was the largest single-stock event of the year. Merck and Moderna reported that their personalised mRNA cancer vaccine combined with Keytruda hit its primary endpoint in a Phase 3 melanoma trial of more than 1,100 post-surgical patients — a 49% reduction in the risk of recurrence or death and a 59% reduction in the risk of distant metastasis or death against Keytruda alone, the first Phase 3 success for an mRNA cancer therapy. Moderna closed up 176.97% at $174.38; Merck closed up 12.60% at $152.20 and carried the Dow. Rieder’s desk started work on the tools businesses the same day, and Newton flagged Merck’s breakout to new highs on heavy volume. The read-through is not the two tickers. It is that a category the market had written down as unprovable now has a Phase 3 endpoint.

Overnight, Asia was watching something else. The Kospi rose 5.89% and triggered a circuit-breaker on capital returns rather than AI demand — SK Hynix up 12.73% on an accelerated buyback and cancellation programme, Samsung up 9.49% on its own. The Nikkei added over a percent and the Hang Seng about 1.3%; European futures were marginally softer. That is a corporate-action bounce, not a demand signal, which matters because the US chip complex has now sold off two sessions running.

Three items for the log rather than the argument. OpenAI’s finance chief told employees the company will be public in 2027 or sooner if the business inflects — a listing timetable, not a valuation event, but one that will shape private-mark comparisons all autumn. Crypto-linked equities rallied on a White House industry meeting attended by the SEC and CFTC chairs. And Tom Lee’s only recoverable equity call inside the window reached the air without him: a Fundstrat idea to avoid Robinhood in 2026, discussed by a panel he was not on, with no price target or rating published alongside it.

Three people voted to raise, and the front end believed them

Wednesday afternoon’s minutes settled the shape of the debate. The July 28–29 meeting held the target range at 3.50–3.75% on a 9–3 vote, with all three dissents preferring a twenty-five basis point increase. Nobody dissented toward easing. That is why today’s claims number carries more weight than its tier suggests: there is no dovish bloc on this committee to reveal, so relative to the vote, the only direction a document or a data point can surprise in is hawkish. The minutes also recorded Chair Warsh raising the idea of cutting the calendar from eight meetings a year to six, letting more information accumulate between decisions. No decision was taken and the 2026 schedule is unchanged.

Here is the number that deserves more attention than it is getting. The front end of the Treasury curve is priced roughly forty-five basis points above the top of the policy range — a market that expects the next move to be up, and to be more than a single hike. September pricing agrees: a hold is the base case at roughly two-thirds, a hike carries most of the remainder, a cut is a small residual. Those odds swung hard this summer, running as high as four-in-five for a hike a few weeks ago before a weak July employment report and Wednesday’s yield relief pulled them back. Treat the exact percentage as directional.

Against that, the most experienced allocator to speak this week said the opposite. Rieder’s view is that the first move Warsh makes is a cut, next year. Goldman’s chief economist published on Sunday that markets are too hawkish on a hike and that the house baseline is on hold through 2026 — a note that landed before the buyback and has not been restated. So the most-followed views on both sides of the Street sit on the dovish side of a front end priced for tightening. Somebody is wrong by the better part of seventy-five basis points, and the instrument that resolves it is not the thirty-year everyone spent Wednesday trading.

The calendar gives this five sessions to marinate. No Fed speakers today and none of consequence before Jackson Hole on August 27–29, where Warsh delivers his first keynote as chair on the Friday. He has signalled the speech will address long-term structural questions rather than near-term guidance, and has made a point of not being constrained by market prices; survey work circulating this week has a majority of managers expecting a neutral tone. If that holds, the front end’s forty-five basis points will not be resolved in Wyoming either — which leaves the data, starting at 8:30 this morning. The one genuine catalyst nobody is positioned for is the thing Rieder asked for: a published reaction function, naming which indicators and which employment series would change the committee’s mind. That would be the most consequential communication change of the year, and it is not what the market is listening for.