Markers Wednesday: X3 Passed on Both Legs, and the Bond Market Handed Us a Cleaner Divergence Test

This column writes falsifiable markers in public, dates them, and grades them whether or not the answer flatters us. Today three of them moved and one of them is unanswerable until this afternoon. We will take those in the order the evidence arrived, and then we will use the week’s mainstream story — a global rout in long-dated government bonds — to write the sharpest divergence test we have put on the board.

X3: PASS, and it was not close

The bar we set on Tuesday was deliberately low: at least one of the August 17 and August 18 sessions prints a positive net flow into the US spot bitcoin ETF complex. We set it low because the preceding week had been ugly and because a low bar that fails tells you more than a high bar that fails. It passed on both legs. Monday: +$297.5 million. Tuesday: +$189.3 million. Two-day total $486.8 million (Farside Investors, retrieved directly at 06:10 UTC Wednesday).

The composition is worth as much as the total. BlackRock’s IBIT supplied $160.2 million then $143.6 million — 62.4% of the two-day figure — which is roughly its long-run share of the complex and therefore tells us nothing unusual. What does tell us something is the breadth on Tuesday: IBIT, FBTC, BITB, ARKB and the Grayscale mini-trust all positive, with only VanEck’s HODL negative at −$16.9 million. Monday was four funds; Tuesday was six. A one-fund inflow is a single allocator rebalancing. A six-fund inflow across two sessions is a distribution channel switching direction.

Now the deflator, because this is where flow analysis usually goes wrong. $486.8 million of net creations does not mean $486.8 million of new conviction. Authorised participants create shares to fill orders that already exist; a chunk of any two-day burst is basis-trade re-establishment, model-portfolio rebalancing on a calendar schedule, and tax-lot housekeeping that has no view on bitcoin at all. The honest claim is narrower and still useful: the five-session bleed from August 10 to August 14, which totalled −$385.2 million and was the largest weekly withdrawal since June, has been fully reversed and then some. Over the seven sessions the complex is +$101.6 million net. The distribution channel stopped selling. That is all we can say, and it is more than we could say on Monday.

A methodology finding we are adding to the house rules

Monday’s session was published around the market as +$137.3 million before it was published as +$297.5 million, and both numbers came from the same tracker. The difference is entirely BlackRock: at the time the first figure circulated, the IBIT cell was blank — unavailable, not zero — and the sum of FBTC ($111.9m), ARKB ($14.2m) and MSBT ($11.2m) is exactly $137.3 million. When IBIT’s $160.2 million landed, the session completed at $297.5 million. Several outlets then filed the completed total under Tuesday, August 18, which is how one session becomes two in a reader’s notes.

The rule we are adopting from this: never grade a flow marker on a total whose largest single component is unpopulated. IBIT alone has averaged $94.2 million a day across the life of the complex; a total computed without it is not a smaller version of the right answer, it is a different question. This sits alongside the range-citing standard we adopted when trackers began disagreeing by more than a million dollars on daily prints. Where sources conflict, print both. Where a source is incomplete, wait.

U2: the projection held its sign, and gave a little back

For a fourth consecutive morning we pulled the mempool.space difficulty-adjustment endpoint at approximately 06:10 UTC. The series now reads −1.63% (Sunday), −0.28% (Monday), +0.49% (Tuesday), +0.37% (Wednesday). The period is 74.80% elapsed with 508 blocks to height 963,648 and an estimated settlement of approximately Saturday, August 22. The three-day average hashrate printed 937.7 EH/s.

U2’s bar is a settled retarget at or above 0.00%, and on the evidence it is on track — but we repeat the caveat we printed when we wrote it, because it is the whole methodological point. This projection is not a forecast of hashrate. It is an extrapolation of the period-to-date average block interval onto the blocks that have not been mined yet. With a quarter of the period left, 508 blocks of unusual luck in either direction can still move it across zero, and a twelve-basis-point overnight retreat is exactly the size of wobble that noise produces. What makes the series interesting is not any single reading but the 2.00-point move from Sunday to Wednesday with a change of sign in the middle. Miners have been adding machines into a $63–65k price and a week of ETF outflows. Saturday tells us whether they meant it.

U3: the closest kind of miss

U3 asks for a daily close at or above $65,000 on or before Friday, August 28. On Tuesday bitcoin printed a session high of $65,058.81 — through the bar, and its first trade above $65,000 since August 10 — and then closed at $64,725.42. The gap to the bar is now 0.42%, against 2.2% when we wrote the marker on Monday of last week and 0.73% after Monday’s rally.

We are not going to soften this into a partial pass, and the reason is a lesson this board has already paid for. In July we graded a marker on an intraday touch and spent two weeks discovering that intraday highs are the most easily manufactured number on any chart. A close is a settlement; a wick is an argument. U3 asked for a close and got a wick. It remains live with seven sessions to run, and it now sits inside the range of a single ordinary green day.

X1: we cannot grade this yet, and we are not going to pretend

X1 asks whether the minutes of the July 28–29 meeting show that more than the three named dissenters favoured an increase in the target range. The Committee held at 3.50–3.75% on a 9–3 vote, with Beth Hammack, Neel Kashkari and Lorie Logan voting for a quarter-point hike. The minutes publish at 2:00 p.m. ET today — roughly twelve hours after this article goes out. We flag it here rather than quietly omitting it, because a marker board that only reports on the days it has answers is a marketing device.

The specific thing to read for is the quantifier ladder in the participants’ section. “A few” is smaller than “several,” which is smaller than “some,” which is smaller than “many,” which is smaller than “most.” The dissent count tells you who voted; the quantifiers tell you who argued. A formulation along the lines of “several participants judged that a further increase in the target range would likely be appropriate at the next meeting” would pass X1 even though only three names appear on the vote — and it would mean the September meeting is far more contested than a 9–3 tally implies.

The divergence question, and why we are writing a marker instead of a headline

Here is the observation that made this a two-story day. On Tuesday the 30-year US Treasury yield topped 5.33%, its highest since 2007 (feeds put the peak between 5.327% and 5.337%; it closed at 5.285%). Japan’s 10-year touched 2.945%, its highest since September 1996. Germany’s 30-year reached its highest since 2011, France’s since 2008. The S&P 500 fell for a third straight session, down 0.69% to 7,691.76; the Nasdaq fell 1.33%. And bitcoin closed up 0.30%, having spent the prior session up 2.59%. Measured from Friday’s closes: bitcoin +2.67%, S&P 500 −1.21%, Nasdaq −1.92%.

The tempting story writes itself — sovereign debt is repricing, bitcoin is the escape valve, the debasement trade is back. We are not going to print that, for three reasons. First, two sessions is not a regime; bitcoin has diverged from equities for two days at a time dozens of times and reconverged within a week on almost all of them. Second, the debasement trade has been an unreliable frame all year: gold is down about 21.5% from its January 2026 peak near $5,600, trading $4,394.08 on Tuesday, and silver is down about 46.6% from its all-time high of $121.62 set on January 29, 2026, at $65.00 — and the fiscal facts underneath that trade did not improve while it unwound — which suggests the trade was crowded and rate-sensitive rather than correct. Third, and most awkwardly for the bullish read, Strategy’s shares fell 5.28% to $92.52 on Tuesday while spot bitcoin rose. The most rate-sensitive listed expression of bitcoin exposure went down on the day the long bond went up. If higher yields were bullish for bitcoin as a monetary asset, someone forgot to tell the leveraged proxy.

So instead of a thesis we are writing a test. Y2 requires both legs to land: bitcoin’s Friday, August 21 daily close above its August 14 close of $63,043.56, and the S&P 500’s August 21 close below its August 14 close of 7,785.76. Bitcoin up while equities are down, across a full week, priced on settled closes rather than wicks. If it passes, we have one clean weekly observation of the divergence and we will say so without extrapolating from it. If either leg fails, the two-day pattern was noise and we will print that instead. Grades Monday, August 24.

New markers

• • Y1 — the US 30-year Treasury yield posts a daily close at or above 5.35% on or before Friday, August 28. Tuesday’s close was 5.285% and Monday’s 5.309%; the intraday peak was quoted between 5.327% and 5.337% depending on the feed. This asks the long end to settle where it has so far only traded, and it deliberately sizes the window to Chair Warsh’s Jackson Hole keynote on August 28. Deadline type: catalyst-sized.
  • Y2 — both legs, as above: bitcoin’s August 21 close above $63,043.56 and the S&P 500’s August 21 close below 7,785.76. Deadline type: scheduled — the week ends when it ends. Grades Monday, August 24.
  • Y3 — total US public debt outstanding crosses $40 trillion in a Treasury Daily Statement dated on or before Friday, August 28. Total public debt outstanding was $39,986,657,878,071.92 on August 18 — $13 billion short — and reporting expects the crossing within days; we write it anyway, because a marker whose failure would be genuinely surprising is the cheapest possible calibration check on our own sourcing. Deadline type: catalyst-sized.

What would change our mind

On the flow story: a single session above −$200 million this week would tell us that Monday and Tuesday were a basis-trade artefact rather than a channel turning, and we would retract the “stopped selling” framing rather than defend it. On the divergence story: a day when the 30-year yield falls and bitcoin falls with it would be worth more than a week of the pattern holding, because it would show the two are moving on a common risk factor rather than on the monetary story. On U2: a settled retarget below zero on Saturday would mean four consecutive daily projections pointed at the wrong answer, and we would publish the post-mortem on our own extrapolation rather than on the miners.

One item landed after Tuesday’s close that changes the policy backdrop this board has been carrying: the SEC proposed Regulation Crypto Assets on Tuesday, August 18 — without an open meeting, five days after cancelling the meeting at which it was to be voted. Any reading of last week that treated the cancellation as a shelving was wrong, and today’s Update carries the detail. And on X1: if the minutes show no support beyond the three dissenters — if the quantifier is “a couple” or the discussion is confined to a named minority — then the hawkish tilt this desk has read into July was overstated, and the September meeting is less contested than the 9–3 vote made it look. We will grade it tomorrow, with the sentence quoted in full.

Did the bitcoin ETF outflow streak end in August 2026?

Yes. The streak ended at three consecutive red sessions. US spot bitcoin ETFs recorded +$297.5 million on Monday, August 17 and +$189.3 million on Tuesday, August 18, totalling $486.8 million and more than reversing the −$385.2 million net outflow of the previous five sessions.

What is the projected bitcoin difficulty adjustment for August 22, 2026?

As of 06:09 UTC on August 19, mempool.space projected +0.37% for the retarget at block 963,648, with the period 74.80% elapsed and 508 blocks remaining. Our own snapshots at the same hour on the preceding three days read −1.63%, −0.28% and +0.49%. The projection extrapolates the period-to-date average block interval and is not a hashrate forecast.

Did bitcoin close above $65,000 in August 2026?

Not as of August 18. Bitcoin traded as high as $65,058.81 on Tuesday, August 18 — its first tick above $65,000 since August 10 — but closed the session at $64,725.42, leaving it 0.42% below the $65,000 line.

Is bitcoin diverging from stocks during the 2026 bond selloff?

Over the two sessions from August 14 to August 18, bitcoin gained 2.67% while the S&P 500 lost 1.21% and the Nasdaq lost 1.92%. Two sessions is not a regime, and Strategy’s shares — the most rate-sensitive listed bitcoin proxy — fell 5.28% on Tuesday while spot bitcoin rose, which cuts against the simple monetary-hedge reading.

When are the July 2026 FOMC minutes released?

At 2:00 p.m. ET on Wednesday, August 19, 2026, covering the July 28–29 meeting at which the Committee held rates at 3.50–3.75% on a 9–3 vote with Hammack, Kashkari and Logan dissenting in favour of a quarter-point increase.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Cryptocurrencies, bonds and crypto-linked equities are volatile and you can lose money. Do your own research and consult a licensed financial advisor before making investment decisions.