The ten-year Treasury note closed on Tuesday at 5.26%. The last time it closed higher was 12 June 2007. The thirty-year closed at 5.59%, and for that one you have to go back to 14 May 2004.

Strip out the inflation story and it gets starker. The ten-year real yield, on Treasury's own par real curve, was 2.91% on Tuesday. Over thirty days the real yield has risen 49 basis points and the ten-year breakeven, which is the bond market's own inflation forecast, has moved 3. Almost the whole rise in what the government pays to borrow is the price of money itself, and not a view about prices.

That is the headwind, and it points straight at anything that pays you nothing to hold it.

Credit repriced through it, and it did so in a way that rules out the easy explanation. ICE BofA's US high yield index carried an option-adjusted spread of 2.66% on 21 September. On Monday it printed 3.02%. Thirty-six basis points in five sessions, and 42 off the 2026 low of 2.60% set on 28 August. CCC and lower went 10.77% to 11.46%. BB, which is where most of the money in that index actually sits, went 1.53% to 1.83%, so this is not one distressed borrower dragging the bottom of the pile.

Here is the objection, and it is the right one to make. High yield cash bonds mark slowly, so a Treasury selloff can turn up in the spread a few days later as an artefact of measurement rather than as a decision anybody took. Fine. Then look at investment grade across exactly the same five sessions: 0.77% to 0.83%. Six basis points. Investment grade carries roughly twice the duration of high yield, so a purely mechanical rates move would have hit it harder, not a sixth as hard. It did the opposite of what the artefact story needs it to do.

And then there is Friday.

On 25 September the S&P 500 rose 0.51% to 7,743.41 and the Nasdaq rose 0.48%. The VIX fell from 15.67 to 14.87. The ten-year fell a basis point. High yield spreads widened 13 basis points, the worst single day of the entire move. Equities buying, volatility falling, Treasuries flat, and credit having its worst afternoon in months. That is not a rates artefact and it is not a general risk-off, because on Friday there was no risk-off to join.

Monday was the risk-off. The S&P fell 0.77%, the Nasdaq 0.92%, the VIX went to 16.07, the ten-year added seven basis points and credit added another nine. So the sequence reads: a rates shock front-loaded on the 23rd and 24th, something that repriced credit and only credit on the 25th, then the broad market catching up on the 28th. Rates went first. They did not mechanically produce what followed.

Nothing broke, either, which is worth saying because a 36 basis point week invites the assumption that something did. HYG closed Monday at a 0.37% premium to net asset value, $77.54 against $77.2563, on a 30-day median bid-ask of 0.01%, from iShares' own page, where those are 28 September figures and a later close will be showing by the time you click. A fund run shows up as a discount and a wide spread. This was orderly, and orderly is the more interesting outcome, because it means somebody sat down and decided it.

Gold did what the textbook says. The LBMA afternoon fix fell 2.73% on Monday, $4,261.05 to $4,144.55, its largest one-day drop on that fix since 1 September. Over the 31 days from 28 August to 28 September, gold is down 9.17% and bitcoin is up 7.22%. That is a split of 16.4 points, against the just-under-fifteen we printed on Monday for the four weeks to the 25th.

Bitcoin, meanwhile, did nothing at all. Coinbase daily closes since Friday run 84,093, 84,417, 84,462 and 83,457, and it was about $83,654 as this was written. Across the whole window since Monday morning it is up roughly 0.7%. The cost of money hit a nineteen-year high and the asset whose entire bull case is a falling discount rate moved less than one per cent in either direction.

It is also on course for its best quarter since 2024. From the 30 June close of $58,523.93 to Monday's $83,456.74, bitcoin has added 42.60%, after three consecutive losing quarters of minus 23.29%, minus 22.03% and minus 14.22%. The best since Q4 2024, at plus 47.48%. The quarter does not close until 23:59 UTC on Wednesday, after this email, so that is a position and not a result, and two sessions can still take it away.

We are not going to resolve this for you, because we cannot. One of those two assets is wrong about what a 2.91% real yield means. The textbook says it is bitcoin, and the textbook has been saying so for three months while bitcoin added 42%.

Three metrics.

Metric 1, Fear and Greed: 73 on Tuesday. Seventy-four on Monday, 71 on Friday, and a 70 to 74 band for six days. Greed throughout, and a band that narrow is not a regime change in either direction. The gauge has spent a week doing nothing while the bond market did a great deal.

Metric 2, ETF flows: plus $31.0 million on Monday, the sixth consecutive session of decay. The run reads 999.0, 714.7, 346.9, 190.7, 134.5 and 31.0. Each session is roughly half the one before it, and every single one is positive, which is harder to describe honestly than either a streak or an outflow would be. The first three are plus $2,060.6 million between them, the three since are plus $356.2 million, and eight of the twelve funds printed exactly zero on Monday. September stands at plus $2,730.2 million against August's final $3,539.1 million, which makes it 2026's second-best month with two sessions to run. One thing not to misread: the complex's net assets fell from $110.84 billion on 22 September to $107.82 billion on Monday while flows were positive every day in between. That is the price, not redemptions. Tuesday's figure is unpublished by either tracker as we write, and Farside's 29 September row, showing a total of 0.0 with every issuer dashed, is a placeholder rather than a flat day.

Metric 3, perpetual funding: 0.00333%, which is zero to two decimal places. That is Deribit's eight-hour rate on the bitcoin perpetual, against 0.000003% when we printed it on Monday. Neither figure is anybody paying up to be positioned, in either direction, at any size a reader of this is likely to carry. Which is this whole issue in two lines. Borrowing dollars got 49 basis points more expensive in a month. Borrowing bitcoin is free.

The STACK retires a published rule this morning, and it is not a rule that lost. It is a rule that could never have been taken. Rule, number, verdict and consequence, in that order, down there.

📣 THIS SPOT IS OPEN

Baseline Crypto reaches nearly 7,000 Bitcoin-focused readers every Monday and Wednesday morning, with an open rate just under 50% over the last four weeks. Direct sponsor placements are open, premium position, top of the issue, one per send.

Reply to this email for rates.

There is no squeeze. The facility the Fed built to catch one is lending one to four million dollars a day, it is lending less of it with each day into quarter end, and the reason nobody touches it is that it is priced above the worst trade in the market.

The premise is not silly and the dates are right. Wednesday is quarter end, month end and fiscal year end at once. It is the classic moment for overnight funding to gap, and September 2019 happened on a date exactly like it. SOFR really is firmer: 3.85% on 21 September, then 3.87, 3.87, 3.88, 3.90 on the 25th and 3.90 again on Monday. Five basis points on the week. The reverse repo facility really did take $11.446 billion on Tuesday against a fraction of that the day before. And the calendar really is heavy: $202 billion of coupon debt settles on Wednesday across reopened ten-year TIPS, twos, fives and sevens, with bill sizes going up on top of it, the seventeen-week to $75 billion auctioning at half past eleven on Wednesday morning. Two hundred billion dollars settling on the one day of the year when every balance sheet is being photographed is exactly the setup a squeeze would need. Anyone pointing at all of that is pointing at real numbers.

So take the evidence one piece at a time.

SOFR is not above interest on reserves. It is exactly on it. IORB is 3.90%. SOFR is 3.90%. The spread is zero, which puts overnight funding ten basis points below the top of the target range and ten below the standing repo facility rate. And "spiked" is the wrong verb for how it got there. It reached 3.90% on the 25th, not on Monday, and the path was 3.85, 3.87, 3.87, 3.88, 3.90, 3.90. That is a grind.

The step everyone is pointing at is the Fed's own rate rise. SOFR went 3.62% to 3.85% on 17 September. Twenty-three basis points overnight, which on a thirty-day chart looks precisely like a funding event. It is the 16 September increase arriving in the overnight market the following morning. It is the single most misread line in this entire story, and it is misread because the chart does not label it.

The facility built for exactly this is empty. Standing repo facility take-up is running at one to four million dollars a day into quarter end, against a SOFR market that traded $2.964 trillion on Monday. It is lower than it was in mid-September. Every bid has been filled in full and nobody has been rationed. Stress looks like take-up that is high and rising. This is negligible and falling.

The mechanism is a price, and it is what explains an empty facility rather than a broken one. The standing repo facility lends at 4.00%. SOFR's 99th percentile on Monday printed 3.98%. So the single worst-priced trade in a market of nearly three trillion dollars still cleared two basis points cheaper than the Fed's backstop. Nobody is being turned away from the facility. Everybody has somewhere better to go.

The reverse repo jump is the easiest of the three. $11.446 billion across ten counterparties is a thirteenfold move on the day before, and against a cap of $160 billion per counterparty it is a rounding error either way. Money market funds tidy their balance sheets for a quarter-end photograph and untidy them the next morning. That is the entire event.

Then there is what Treasury itself did on Tuesday, which nobody running the squeeze story has mentioned. It ran a TIPS buyback against a $750 million maximum, dealers offered $3.981 billion, and it accepted $605 million. An issuer worried about its own market does not walk away from $145 million of capacity it was entitled to use on the eve of quarter end because it did not like the prices.

The Desk has said the rest of it out loud, though you have to reach one day outside our window to find it. On 22 September the System Open Market Account manager, Roberto Perli, said the Desk has paused its reserve-management bill purchases for two purchase periods because it judges reserves ample, and that roughly $400 billion of net bill issuance produced only "very modest upward pressure on repo rates."

Now ask what that man is paid to avoid. A SOMA manager who lets the funding market gap owns a permanent failure with his name on it, in the same way a Treasury Secretary who runs the cash account too low does. Nobody in that chair under-reacts to a reserve shortage. The incentive runs entirely the other way, toward buying bills and saying nothing. He stopped buying bills and said so out loud.

One absence is worth noticing too. Five Fed voters published prepared texts inside this window and not one of them mentions repo, funding or quarter end. Cyber risk, community banks, AI, payments and the policy rate. When the plumbing genuinely worries that building, the building talks about the plumbing.

What we are not claiming. We are not saying Wednesday is a non-event. Overnight rates do print firmer over quarter end, they are supposed to, and a few basis points on 30 September would be ordinary rather than informative. What we are saying is narrower: the specific claim that the system is short of reserves and about to show it is not supported by any of the numbers being offered for it. It is also checkable within hours. Tuesday's SOFR publishes at eight this morning, half an hour after this email. The quarter-end reverse repo operation prints about a quarter past one this afternoon. Wednesday's SOFR publishes at eight tomorrow.

New dated test, grading Monday 5 October. The number is the Secured Overnight Financing Rate published by the New York Fed for 30 September 2026, which posts on 1 October. Branch A: 4.00% or above, meaning overnight funding cleared at or through the top of the target range on the one date this argument says it will not, and this section is wrong. Branch B: 3.90% up to but not including 4.00%, which is where it has printed since the hike and is what the argument above implies. Branch C: 3.80% up to but not including 3.90%. Branch D: below 3.80%. One published series, one date, four ranges, no gap between them and nothing outside them. Branch A is the branch that costs us, and it is named first for that reason.

Three numbers that define the next week.

2.91%, which is already past the top we called on Monday. On Monday we printed the ten-year real yield at 2.85% on Thursday 24 September and called it the highest since 2008. It has been past that twice since. Treasury's own par real curve reads 2.90% for Monday and 2.91% for Tuesday, both fresh post-2008 highs, and both above the level we had just described as the peak of the move. It is still under the 2008 high of 3.15%. One precision note, because this claim gets stretched: Treasury's par real series begins on 2 January 2003, so "highest since 2008" is supportable inside that series and nowhere earlier, and we are not extending it. Next to it, the ten-year breakeven has moved three basis points in thirty days while the real yield moved forty-nine. The bond market has not changed its mind about inflation. It has changed its mind about what money costs. If you hold something that pays you nothing, that second sentence is the only one that matters, and it has moved half a percentage point against you inside a month. Our 14 September test takes its reading on Wednesday and grades on 7 October. It names FRED's DFII10, which reads 2.90% for 28 September because FRED is running a session behind Treasury, and it grades on FRED's print. We are not swapping series at grading time.

08:31 on Wednesday morning, which is one minute after the number everybody will be comparing to. August PCE lands at half past eight, an hour after this email, with the third estimate of Q2 GDP and corporate profits in the same minute. That much is ordinary. What is not ordinary is that Wednesday is also the day BEA begins its 2026 annual update, and in its own words, "the 2026 annual updates of national, industry, and regional data will begin on the same day for the first time: September 30, 2026." The revision runs from 2021:Q1 through 2026:Q1 and takes in monthly personal income and outlays along with everything else. Which means July's PCE figures can themselves change at 08:31, and August's year-on-year will be computed off a base that may not be the base you remember. If you see "core accelerated from 3.3%" on Wednesday morning, check whether the 3.3% survived the revision before you believe the acceleration. The Cleveland Fed's nowcast, as it read on Tuesday, points to core at 3.4%. It re-runs every business day, so the page will have moved on by the time you open it, and it is a model built on the current vintage, so it does not anticipate the revision either. The line in that release we would actually read is corporate profits, arriving on the same morning that credit has spent a week repricing the risk that companies do not pay.

121,172 coins on 30 October, and one that tripled and then vanished before you read this. Deribit's total open interest is 361,336 bitcoin, and the 30 October expiry carries 121,172 of them at a 0.33 put-to-call ratio, about $10.1 billion of notional. On Monday we printed 118,263 at 0.31, so it has put on 2,909 coins in two days and got fractionally less call-heavy doing it. It expires two days after the October FOMC, which is what makes it the one worth watching rather than this week's. This week's earns a sentence for a different reason. The 30 September daily we described on Monday at 2,225 coins was carrying 7,132 last night, more than triple in two days as the date came at it, and it settled at 08:00 UTC, four o'clock this morning, three and a half hours before this email. Open interest built into a date and then expired before anybody had read a word about it.

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. A rule we published on 2 September fired on Monday, and there was nothing to execute. Rule, number, verdict, consequence, in that order.

The rule, published in this section on 2 September, in full: "The entry condition stands: the next position comes from the reserve drain thesis or from nothing, and if high yield spreads widen through 3.00% into the October peak, that is the trade, published here before it is taken." We restated it on 9 September as the trade we had published and not taken, and we have carried it here in every issue since.

The number: 3.02%. ICE BofA US High Yield OAS, FRED series BAMLH0A0HYM2, for 28 September, against 2.66% on the 21st. It is the highest since 7 April. It is not the first time in this cycle, and we nearly wrote that it was: spreads sat above 3.00% through most of March. That does not weaken anything, because the rule was written on 2 September with spreads in the 2.6s. The line it named was crossed.

The verdict: the rule fired, and there was nothing to execute, because the rule never named an instrument. Read the sentence again. If spreads widen through 3.00%, that is the trade. It names a level and a window and nothing else. No contract, no ticker, no direction, no size, no exit. The clause on the end, published here before it is taken, deferred all of that to an issue that never came, because the level never printed until Monday. We have been back through all seven issues the rule ran in and it names an instrument in none of them. It read like discipline for four weeks and it committed to nothing.

One thing we are not going to bury inside that. Monday's issue told you this line was 20 basis points away. That was written off the 24 September value of 2.80%, the latest FRED had at two in the morning, and the 25 September print of 2.93% published later that day. So on the morning we said twenty, it was seven, and by that evening it had crossed. No goalpost moved. The series lags, and you should know by how much.

The consequence: it is retired, in print, this morning. The same way we retired the CLARITY re-entry rule on 16 September, and for a cleaner reason. We are not replacing it with a position today. Reverse-engineering a trade after the trigger has already fired is exactly the goalpost-moving this section grades other people for, and it would be worth less than nothing.

The standing rule that comes out of it, and it binds everything published in this section from here. An entry condition has to name what we would buy or sell, at what level, and what takes us out. If that sentence cannot be written, it is an observation, and it gets published as an observation.

And one thing that does not grade today. The separate three-way commitment we published on 2 September is scoped to the moment the Treasury cash balance peaks, in late October. Monday's print does not settle it either way.

The four open tests, with live levels.

High yield spreads on 30 September, grading 7 October. Branch A is 3.00% or above, and Monday's print puts it there, against Branch B in Monday's issue. It takes its reading on Wednesday and it does not grade today. When we set this on 9 September the latest value was 2.68%, sitting in Branch C, and we wrote that Branch A was what our own argument implied and 32 basis points away.

The ten-year real yield on 30 September, grading 7 October. Branch A is 2.80% or above. Live: 2.90% on FRED's print for 28 September, and 2.91% on Treasury's own curve for Tuesday. Both are Branch A. The rule names FRED and grades on FRED. Treasury is here for the live level and nothing else.

US spot bitcoin ETFs net positive for September, grading 7 October. Tracking Branch A at plus $2,730.2 million through Monday, on our own sum of Farside's printed daily totals, which is the measure we stated in print. Flipping the month negative in the two sessions left now needs two consecutive outflows each worse than the minus $1,113.7 million record. On Monday it took three merely enormous days.

The Treasury cash close on 30 September, grading 5 October. Branch C is below $950 billion and Branch B runs $950 billion to $1,050 billion. Live: $959.572 billion on Monday's Daily Treasury Statement, which is $9.572 billion over the line and into Branch B. On Monday we printed $924.627 billion and called it $25.4 billion under. The two statements since have added $34.945 billion between them. This was comfortably Branch C seventy-two hours ago and it is tracking B now, with one business day to run. The close that settles it is Wednesday's, and it does not publish until Thursday afternoon, so this is a number nobody can know at half past seven on Wednesday morning, us included.

Fed decision in October, 25 bps increase | Market: 44.5% 🟡 NO TRADE, and our own rule is closer than it has ever been
Levels here are Tuesday evening. No change is 53.5%. On Monday we printed 64.5% and said the entry might never trigger before the meeting. It peaked at 70.5% that same day. Then at two o'clock on Tuesday afternoon New York Fed President John Williams told an audience in Buffalo that "there is no need for urgency, and we have time to gather more information," and inside fifteen minutes this leg went 67.5% to 50.5%. Governor Barr had spoken at the Detroit Economic Club at lunchtime and said that "in my base case, further policy adjustments are likely to be needed to ensure inflation comes down to target in a timely fashion," which is word for word what he said in the housing speech we quoted on Monday, six days earlier. The market gave a verbatim repeat exactly the weight it deserved, which was none. Peak to now is 26 points. The reason to trust it is what happened after: 47.5% on every fifteen-minute bar from three o'clock until four, the two legs summing to about 98% and each carrying $4.3 to $4.7 million of volume, and then 48.5, 46.5, 45.5 and 44.5 by twenty to five. A twenty-point move that snaps back inside the hour is a thin market. One that gaps, sits still for an hour and then keeps going is a repricing the market has had time to argue with and has not taken back. What would put us in: this leg below 40%. That was 24.5 points away on Monday morning. It is 4.5 points away now, and it is a rule that names a contract, a side and a price, which is more than the one retired above ever did.

Fed decision in December, 25 bps increase | Market: 75.5% 🟡 NO TRADE, and the tell is what this one did not do
Here is the test that kills the obvious story. If Tuesday had been the market rescheduling a hike rather than removing one, December would have picked up what October dropped. December went 76.5% to 75.5% while October fell twenty-three. It lost a single point. The hike was priced out of the front end, not moved down the calendar. Over the week the two have gone in opposite directions: we printed December at 67.5% on Monday, so it is up eight points since, while October is down twenty. That is a market growing more confident about the destination and much less confident about the timing. What would move it: the October decision itself on 28 October, which settles the timing question in an afternoon and this contract five weeks later.

Zero Fed rate cuts in 2026 | Market: 96.25% ⚫ EXITED at 88.75%, a loss, and it came back a point
Short from 78%, trimmed at 85%, covered at 88.75% on a published stop on 3 August, recorded as a loss in every issue since. We printed 97.25% on Monday and it is 96.25% now. That is a point back our way on a position we do not hold, which is worth exactly nothing and gets printed anyway, because a record that only reports the drift when the drift hurts is doing something other than reporting. For the record, and against our own habit: it did go our way for a fortnight after the stop, bottoming at 85.15% on 17 August. What would move it properly: not a soft data print and not a speech. One dot below the current midpoint in the December projections, from anybody in the room.

CLARITY Act signed into law in 2026 | Market: 5.5% ⚪ NO TRADE, and it is the precedent for what happened at the top of this section
5.7% on Monday, 5.5% now, and we hold no live rule here because we retired the re-entry condition in print on 16 September, when the markup leg it turned on became unreachable. That retirement is the reason this morning's is not a novel act. Two published rules have now come off this board inside a fortnight, for two different defects: one whose condition could no longer occur, and one whose condition occurred and meant nothing. Those are the two ways a rule dies, and neither of them is being wrong. What would make this tradeable again: a markup of the substitute text, and a new rule at a new price written before the fact rather than after it. Neither exists.

The book. Flat for the seventeenth consecutive issue, recomputed from the archive rather than carried forward: sixteen issues published after the 3 August exits, so this is seventeen. On Monday we told you two published rules were live. This morning there is one. The October leg below 40%, four and a half points away, on a market where we understand the resolution language and where being wrong stays cheap because the meeting is four weeks out. That is the entire board, and it is smaller and more honest than it was on Monday.

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

On Monday we asked one question with a one-word answer: do you have bitcoin sitting on an exchange this morning, yes or no.

Nothing came back. We went through the inbox and our own file of reader replies before writing this, and there is still nothing new since 3 August. That is five issues running.

We said on Monday that we would keep saying so plainly and would not redesign the question again, so we have not. What follows is a different question in the same shape. If the answer is silence again, we will write this paragraph for a sixth time on Monday and it will be the same paragraph.

This issue has one argument in it. Money got more expensive than it has been since 2007, credit repriced through it inside a week, gold fell 2.73% in a day exactly as the textbook says it should, and bitcoin sat there and is on course for its best quarter in two years. Those two responses cannot both be a considered reading of the same information.

One question: which of them is telling the truth about the next six months? Reply with one word, bonds or bitcoin.

If you have a reason, send the reason. If you only have the word, send the word.

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

See you Monday with what the August PCE print looked like once BEA had revised the base underneath it, the Treasury cash close that grades our 2 September test, and whether Wednesday's crude backfill matched the six numbers we printed today.

The Baseline Crypto Team

Two tools. Free, no account, nothing to install, and both checked again before this went out.

The hardware wallet checker takes your device and firmware version and tells you whether you are sitting 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 it is deliberately unkind about backups and inheritance, because that is where almost everybody actually fails.

You have a price in your head. If bitcoin gets to some number, you will move coins off an exchange, or add, or finally do the thing you have been meaning to do since the spring. You have never written it down as an order. There is no venue in it, no size, and nothing in it that tells you when you were wrong. That is the rule we retired three sections up, in miniature, and a number with no instrument attached is a feeling wearing a decimal point.

So open the custody audit, and before you answer question one, write two sentences at the top of the page. The price at which you would move, and exactly what you would do in the hour afterwards. If you cannot finish the second sentence, the first one was never a plan.

Then send it on to the person you thought of while reading that. You know the one. The friend who has told you twice what price he would sell at and has never once said where.