On Wednesday we promised you three verdicts by this morning. All three came back the same way, and the way they came back is the reason this issue is about something else entirely.

CPI was a non-event. Core printed 2.5% against June's 2.6%, headline 3.4%, both in line. We told you the number to watch was not the print itself but the ten year breakeven, because if it did not move, the bond market had decided the print was noise. It sat at 2.27% and did not twitch.

Credit did not blink. We staked Wednesday's issue on a contradiction: high yield spreads at 2.70% and ninety seven billion dollars leaving the banking system could not both be right. Spreads are now 2.71%. One basis point in five sessions.

And the drain stopped. The Treasury took back $4.9 billion this week against $96.8 billion the week before. The thing we said to watch simply switched itself off.

So every macro question we put in front of you resolved to nothing. Bitcoin fell anyway, from $63,762 to $63,064, drifting down through a week with no bad news in it at all.

When the macro explains nothing, stop asking the macro. Ask who is actually buying. The answer turns out to be strange, and almost nobody has written it down: Strategy's preferred stock funded roughly 77,000 bitcoin of purchases in 2026. Every US spot bitcoin ETF combined took in about 8,000. One company's financing instrument has been a bigger bid than the entire ETF complex, by nearly ten to one.

Three metrics. Here is the read.

Metric 1, Fear and Greed: 31. Up four points from Wednesday's 27, in a week when price fell. Sentiment and price came apart, which usually means the people left in the market are the ones who were never going to sell.

Metric 2, credit spreads: 2.71%. The answer to our own question, and it is an anticlimax. We said credit reprices before equities and long before bitcoin. It had five sessions to react to the largest weekly liquidity withdrawal of the year and it moved one basis point. Either the drain was never the threat, or credit is not the early warning we treated it as.

Metric 3, reverse repo: $0.2 billion. Not a typo and not rounded. The facility that peaked above two trillion dollars now holds two hundred million. It is finished as a shock absorber. What is interesting is that the Treasury stopped draining in the same week the buffer ran out, and we do not know whether that is coincidence or restraint.

📣 THIS SPOT IS OPEN

Baseline Crypto reaches 7,000+ Bitcoin-focused readers every Monday and Wednesday morning. Direct sponsor placements are open — premium position, top of the issue, one per send.

Reply to this email for rates.

Start with the one we earned, because it was a real call and it landed.

On August 10 we set a test whose Branch A required core CPI at or below 2.9% and September no-change still above 60% at Friday's close. Core printed 2.5%. No-change closed at 74.5%. Branch A, cleanly. On Wednesday we went further and said the print was close to decided, that the market had narrowed it to two adjacent tenths, and that the useful tell was whether the ten year breakeven moved. It printed 2.5%, one of the two tenths, and the breakeven did not move at all. That is the call working exactly as described.

Now the one we did not earn.

Wednesday's test required three things for Branch A: reverse repo under $5 billion, the Treasury account rising week on week, and bitcoin failing to reclaim $65,000 by Friday's close. All three are true. RRP is $0.2 billion, the account rose, bitcoin's weekly high was $63,916 and it never came close. On the scorecard that is a pass.

It should not count, and here is why. The condition was “the Treasury account rises week on week.” It rose by $4.9 billion. The week we were actually describing, the one that made the story worth telling, was $96.8 billion. A test that treats those two as the same event is not testing anything. Almost any week satisfies “the account went up.” We wrote a condition loose enough to pass on noise, and then it passed on noise.

So the mechanism never got tested. We claimed a liquidity drain was pressuring bitcoin. This week the drain stopped, bitcoin fell anyway, and credit moved a single basis point. Every possible outcome was consistent with our test, which means our test had no information in it.

The honest read on the underlying claim: weakened, not refuted. Bitcoin did keep bleeding, which is what we said would happen. But it bled in a week when the thing we blamed was absent, so the correlation we were pointing at did the opposite of confirming itself.

What we are changing. Dated tests from here specify magnitude, not direction. “TGA rises” becomes “TGA rises by more than $40 billion.” A condition that cannot fail is not a test, it is decoration, and a scorecard full of decorations is worth less than no scorecard at all.

New dated test, written the way the last one should have been.

Branch A, the bid is thinner than it looks. STRC closes below $100 on Friday August 21, and Strategy discloses a fifth bitcoin sale. Read: the largest single buyer of 2026 is still liquidating to defend its own financing, and the marginal bid stays absent.

Branch B, the bid returns. STRC closes at or above $100 by Friday August 21 with no new bitcoin sale disclosed. Read: Saylor's stated trigger is met on schedule, the seller becomes a buyer again, and the largest overhang in this market lifts.

Graded Wednesday, August 26. Both legs are specific, both can fail, and neither is satisfied by noise.

Three numbers that define the week ahead.

74.5%, September is now essentially closed. No-change has gone from 61.5% on Wednesday to 74.5%, while the 25 basis point hike collapsed from 38.5% to 24.5%. A cut of any size is under 2%. One in-line inflation print did in five sessions what two months of Fed commentary could not, which tells you the September question was always an inflation question wearing a labor-market costume. The useful consequence: rates are no longer the variable. If bitcoin keeps falling from here, you cannot blame the Fed for it, and a lot of commentary is about to try.

$95.77, and the countdown stalled. Strategy's preferred moved twenty two cents on $108.6 million of bitcoin sales. The week before, a similar spend bought more than two dollars. Watch the size of the move per dollar spent rather than the price itself, because that ratio is the real signal about how much selling pressure sits above par. At this rate the $785.2 million remaining does not get there, and Saylor's stated September 8 arithmetic starts to look optimistic rather than conservative.

2.27%, the number that refused to react. The ten year breakeven, unchanged through a CPI print, unchanged for a month. We flagged this on Wednesday as the tell, and it told: the bond market does not think any single monthly print changes the inflation story. Keep watching it, because the day it does move will be far more informative than the day CPI comes in a tenth hot. Inflation expectations moving is a regime change. A CPI print is a data point.

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: two tests graded, one clean pass, one pass we are refusing to bank. The August 10 test resolved Branch A on both legs, core CPI at 2.5% against a 2.9% ceiling and September no-change at 74.5% against a 60% floor. That one was called and it landed. The August 12 test also resolved Branch A on all three legs, and the FILTER above explains at length why we are not counting it: the condition was written loosely enough that noise satisfied it. Recorded as a pass, marked as uninformative. We would rather carry an asterisk than pretend a badly written test taught us something.

Still no position. The book has been flat for four issues. That is not caution as a pose, it is the absence of a price we like. The September contract has moved fourteen points in five sessions and we caught none of it, because our published trigger required the opposite direction. Saying so is cheaper than inventing a reason we were right.

Fed decision in September, 25 bps increase | Market: 24.5% ⚪ NO TRADE, and the question has closed
Down from 38.5% Wednesday and 57.5% two weeks ago. No-change is now 74.5%, a 25 basis point cut sits at 1.2%, and everything else is rounding. One in-line CPI print took fourteen points off the hike. We flagged on Wednesday that three points had drifted back onto the board and suggested the drain might be the cause. It was not. Inflation data was, and it went the other way. Correcting that here rather than leaving it to stand.

Zero Fed rate cuts in all of 2026 | Market: 85.2% ⚫ EXITED, and the exit keeps looking early
We shorted at 78%, covered at 88.75% on a pre-stated rule, and it now trades 85.2%, down another seven tenths. Four issues of slow drift in the direction we originally wanted. The rule that stopped us out was correct as risk management and wrong as a forecast, and both of those stay true no matter how far it drifts. Not re-entering on hindsight.

CLARITY Act signed into law in 2026 | Market: 19.5% ⚫ EXITED, rule keeps validating
Down two more points from Wednesday, now under a fifth and less than half the 41% where the position opened. We closed at 33 to 37% on the stated cloture rule on July 29, before Thune shelved it. Congress returns in September and nothing moves before then. This is the cleanest exit on the board and it keeps getting cleaner.

Retired: Bitcoin dips to $55,000 before 2027. We said on Wednesday we would restate or retire this rather than let it lapse quietly. Retiring it. The entry trigger needed September hike odds above 65% while the contract traded under 60%; hike odds are 24.5% and falling, so the setup cannot occur in any form resembling the original thesis. Rebuilding it around a different variable would be a new idea wearing an old trade's name.

Track the whole board live at polymarket.com, free, no account required.

Wednesday's question, answered: neither, and we asked it wrong.

We asked whether credit was asleep or the drain simply was not binding. The week refused to pick. The drain switched off before credit had to decide anything, so the question never got put to the test. Spreads at 2.71% are not evidence that credit was right. They are evidence that nothing was asked of it.

That is the second time in one issue we have had to say a question we set could not be answered by the week that followed. Both failures share a cause, and it is worth naming: we were writing conditions that described a direction rather than a magnitude. Direction is almost always satisfied. Magnitude is what carries information.

So the new question is simpler and it is about the thing that actually moved.

All year, the biggest bid under bitcoin has not been the ETFs. It has been one company selling preferred stock to fund purchases, roughly ten times what every US spot ETF took in combined. That company is now selling bitcoin, at a loss, to defend the price of the instrument that funds it. The buyer became the seller, and the ETF flows everyone quotes each morning were never the main event.

So: if Strategy stopped issuing tomorrow, what is left holding this market up? Is there a real bid underneath, or have we spent a year mistaking one company's balance sheet for institutional adoption?

If you have modelled this, or you think the 77,000 against 8,000 comparison is unfair for a reason we have missed, we especially want that second one. An argument against our own lead is more useful to us than agreement.

Hit reply. We read every response, and the best answers run Wednesday.

See you Wednesday, with STRC against par, whether a fifth bitcoin sale shows up, and what the ETF complex did in a week when nobody was watching it.

The Baseline Crypto Team

Three tools we built because we wanted them and could not find them. Free, no signup, no accounts, and nothing you type ever leaves your browser.

Is your hardware wallet screwed? Check your device and firmware against every logged advisory, including the ones where updating your firmware does not save you, because the seed itself is already weak.

How fragile is your self-custody? Nine questions, scored out of 27. Every question exists because of a documented loss.

Our track record, losses included. Every dated call we have made, graded in public, including the one we refused to count this morning.

The Coldcard theft has settled near 1,816 bitcoin and the attacker is still working through addresses. If you know someone who bought a Coldcard before 2022 and has been meaning to move their coins since July, send them the first link tonight. They will not do it on their own.