Three markers close today, one settles tomorrow night, and the most interesting number on this desk is not on the marker board at all. Bitcoin is up 19.68% on the week and open interest is lower than it was before the rally started. That combination — a 20% advance the derivatives market did not fund — is the thing worth arguing about, so we grade first and argue second.

W1 — PASS

The bar, written August 15: the CFTC makes an explicit claim to jurisdiction over spot crypto markets on or before August 21. Deadline type: arbitrary, sized to the week Washington had scheduled for itself.

The verdict: at the CFTC's first-ever Innovation Advisory Committee meeting on Thursday, August 20, Chairman Michael Selig told staff to prepare a regulatory framework for crypto markets that does not depend on Congress. His words, as reported by CoinDesk and Bloomberg: "If Clarity continues to stall because of Democrat obstruction, the CFTC will utilize its existing authorities to begin establishing a regime for crypto asset markets."

That is a claim to jurisdiction under existing authority — the specific thing the bar was written to catch, and the thing that had not happened when we wrote it. Six days earlier the SEC's rulemaking vote had been cancelled with eighteen hours' notice and the read on this desk was that no agency wanted to move first. PASS.

Deflators, printed as a matter of course. Selig described a contingency, not a filing: there is no proposed rule, no comment period and no Federal Register entry. A conditional threat is cheaper than a rulemaking, and the conditional is doing real work in that sentence. We graded the claim, which is what the bar asked for — not the follow-through, which is a different marker and not yet written. Full context in today's update.

U1 — FAIL at press time

The bar, written August 17: a further Strategy bitcoin sale disclosed in a filing dated on or before August 21. Deadline type: arbitrary, sized to the company's weekly 8-K rhythm. The thesis behind it was that four sales in four weeks had stopped being a tactic and become a policy.

The verdict: the only filing inside the window landed Monday, August 17, and it disclosed no coin sale at all. Instead: roughly 3.46 million shares issued for $333.7 million, $132.2 million of STRC preferred repurchased, $52.4 million of dividends paid, and $149.1 million added to a US dollar reserve that now stands near $4.8 billion. Holdings remain 840,447 BTC as of August 16, with no purchases and no sales between August 10 and August 16. FAIL.

The same filing carries the week's most quietly remarkable number. Strategy's average cost across those 840,447 coins is $75,385 — a cost basis of $63.357 billion. At Friday morning's $75,447 the position is worth $63.409 billion. After a quarter spent selling coins below cost to fund preferred dividends and buybacks, the largest corporate bitcoin treasury on earth is back above water by roughly $52 million, or seven hundredths of one percent. That is a rounding error, not a victory; but the sign has changed, and the sign is what the equity trades on. MSTR closed Thursday at $112.39, up 7.81%.

We are grading this as a fail rather than holding it open, and stating the caveat plainly: the window technically runs to the end of today, and this desk publishes at roughly 06:30 UTC. If a filing lands in US hours on August 21 disclosing a sale, we will print the correction on Saturday under the same rule we applied to R3 last week — right thesis, wrong deadline, not regraded.

The more useful point is what the failure teaches. Our rule from August 17 was that when a filing shows both a sale and an issuance, the pair is the story. Monday's filing showed an issuance and no sale, which is the same story read from the other end: with the equity lever open, the coins do not have to move. On Thursday Strategy's stock closed at $112.39, up 7.81%. The lever is more open today than it was on Monday. See our field manual for reading these filings.

X3 — PASS, and the floor becomes a number

The bar: US spot bitcoin ETFs record two consecutive positive sessions before the end of the week. Graded PASS on Wednesday with $297.5M and $189.3M. It closes formally today, and the completed rows are worth printing because of what they say about our own caution.

On Thursday morning we published Wednesday's session as a floor of +$164.2M rather than a total, because IBIT's and FBTC's cells were blank — not zero, blank. That was the right call: the completed row prints +$517.2M, with IBIT at $284.7M and FBTC at $62.4M. The floor understated the session by more than three times. Thursday's row is now in at +$606.3M, IBIT $503.0M.

IBIT took 67.8% of the four-day total. Concentration that high is a single-allocator signal as much as a market signal, and it is the reason we keep printing the fund-level table rather than the headline.

U2 — settles tomorrow night, and the projection has flipped twice

The bar, written August 17: the settled difficulty retarget at block 963,648 comes in at or above 0.00%. Deadline type: scheduled, by the protocol itself.

Friday's 06:10 UTC snapshot — the sixth consecutive daily reading from the same mempool.space endpoint at the same hour — prints −1.20%. The full series:

A note on that last row, because precision matters more than tidiness: two calls to the same endpoint inside the same minute returned −1.195% and −1.218%. The endpoint recomputes live off the trailing block times, so a projection quoted to two decimals is spurious at this range. We are printing it as −1.20% and treating anything finer as noise.

With 247 blocks to run, 87.75% of the epoch elapsed, and a trailing average block time of 607,743 milliseconds — 10 minutes 8 seconds against the protocol's 10-minute target — settlement is due around 23:56 UTC on Saturday, August 22. On the current path U2 fails. We will grade it on the settled print and nothing else.

The reason the projection turned is visible in the hashrate series, and it is the counter-signal of the week. Daily average hashrate, same source: 1,025.3 EH/s Monday, 1,000.2 Tuesday, 941.1 Wednesday, 782.1 Thursday, 812.7 Friday (partial). That is a 23.7% decline from Monday's peak, into a 19.68% price rally. Network hashrate now sits roughly 17% below its all-time high and difficulty about 14% below its 2026 peak — only the second year-over-year difficulty decline in the network's history. The dominant explanation in current reporting is structural rather than seasonal: the AI and high-performance-computing pivot, with CoinShares estimating that AI/HPC hosting could reach around 70% of listed-miner revenue by the end of 2026. Summer curtailment in Texas — where large miners cut load during grid-stress hours to avoid the following year's transmission charges under the 4CP program — is real but secondary. Either way, the machines are leaving while the price is rising, which is not the configuration a hashrate-follows-price model predicts. Mechanics in our difficulty adjustment guide.

The find: a 20% rally the derivatives market did not fund

Now the part that is not on the board. Three independent readings from our own Binance futures pull, all taken at 06:13 UTC Friday, say the same thing.

One: open interest never rebuilt. Aggregate BTCUSDT open interest was 111,988 BTC on August 15, before any of this. It fell to 106,273 BTC by August 19 as the squeeze burned positions, and stood at 109,390 BTC on Friday — 2.3% below the pre-rally level, in coin terms, after a 19.68% price move. (Binance's daily history series, sampled at the period boundary rather than live, put Friday's reading at 107,625 BTC, or 3.9% below. Both are correct; we quote the live endpoint and note the other rather than pick the flattering one.) In dollars the notional went the other way, $6.875bn to $7.857bn, purely because the price of each coin-denominated contract went up. Which measure you look at determines what you conclude, and coin-denominated is the honest one for this question.

Two: funding never got excited. The eight-hourly funding rate on the same contract printed 0.0016% on Wednesday morning, peaked at 0.0100% — exactly the venue's baseline — on Wednesday evening and again on Thursday morning, and settled back to 0.0050% at Friday's 00:00 UTC mark. Perpetual funding at or below baseline through a 20% week means longs are not paying a premium to hold. There is no crowding to unwind.

Three: retail flipped short. Binance's global long/short account ratio for BTCUSDT went 2.05 on Saturday, 2.22 on Monday, 1.48 Tuesday, 1.38 Wednesday, 1.08 Thursday and 0.96 on Friday. Retail accounts entered the week better than two-to-one long and are now, in aggregate, net short at the highest price since May.

Two readings, and we hold both. The bullish one: there is no leverage to flush, the marginal buyer this week was a spot ETF wiring $1.61 billion, and a market that rallies without adding leverage is structurally sturdier than one that does. The bearish one: the fuel is spent. Roughly $2.7 billion of shorts — the largest such liquidation in CoinGlass's records, which begin in 2021 — were the buyers of the last 16%, and those buyers are gone. The remaining shorts are a smaller pile than Wednesday's, so the next squeeze is necessarily smaller than the last.

We have written the falsification test rather than the thesis, per the rule adopted August 19. It is Z2 below, and it is the cleanest marker on this board: if leverage rebuilds, the clean rally is over and what is left is a chase.

Live markers, updated

Y1 deserves a word. When we wrote it on Wednesday the 30-year had just closed at 5.194% and the Treasury had announced a buyback expansion explicitly designed to push that number down — we said in print that the marker had become a bet against a government programme and declined to resize it. By Thursday's close the yield was back to 5.237% and it was 5.248% at 02:14 ET Friday. The programme has not started; the operations begin September 9. Markets repriced the announcement in one session and then repriced the reprice.

New markers

Three, all date-anchored, with deadline type stated, per the post-R3 rule.

Z1 — scheduled. Binance BTCUSDT weekly close at 00:00 UTC on Sunday, August 23 is at or above $73,025.15, Thursday's daily close. Bar chosen deliberately: it asks whether the market holds two of the three squeeze days through the first weekend without an ETF bid, not whether it holds the highs. Grades Monday, August 24.

Z2 — arbitrary, sized to Jackson Hole. Binance BTCUSDT aggregate open interest closes at or above 111,988 BTC — the August 15 pre-rally level — on any daily reading on or before Friday, August 28. A PASS falsifies the de-levered-rally thesis above. A FAIL confirms that this move was bought, not borrowed. We would rather be wrong on this one publicly than right quietly.

Z3 — scheduled. The sum of US spot bitcoin ETF net flows across the four sessions August 24–27 inclusive is at or above zero, on Farside's completed rows. Graded on completed rows only, never on a total whose largest component is blank.

The week in one sentence

A regulator claimed a lane, a treasury company chose its equity over its coins, four ETF sessions bought $1.61 billion, roughly $2.7 billion of shorts were carried out — and the derivatives market, which normally writes the ending to weeks like this, sat the whole thing out.

Full context on the tape in today's news report, the framework in How to Read a Short Squeeze, and every field guide this desk has published in the Reading Room.