This desk publishes dated, falsifiable markers and grades them in public whether they pass or fail. Monday settles one, moves two, and opens a third. It also produces the cleanest reading yet of the structural claim we have made every day of this rally — and, less comfortably, the clearest evidence yet that our best marker on the board is the one designed to prove us wrong.

Z1 — PASS, and a candid word about the design

The marker, as written on 21 August: the settled Binance BTCUSDT weekly close for the week beginning 17 August will be at or above $73,025.15. Deadline type: scheduled, 23 August 23:59:59 UTC.

Result: PASS. The week settled at $77,734.00, clearing the bar by $4,708.85, or 6.45%.

Now grade the marker, not just the outcome, because a board that only reports passes is a marketing document. Z1 was written on Friday with bitcoin at roughly $75,447, giving it a cushion of about 3.2% at the moment of writing. Over the two sessions of the window the closest the market came to the bar was Sunday’s low of $75,545.67 — still 3.45% above it. So Z1 was never inside a single day’s ordinary range of failing.

That said, it was not the worthless marker Y2 was. Three sessions this month printed high-to-low ranges of 4% or more (19, 20 and 21 August, at 9.01%, 6.49% and 8.86%), against a next-widest of 2.96% across the other twenty days. One more day of that character inside the window would have tested the bar. Call Z1 fair: gradeable, plausibly testable, never actually tested. The honest summary is that we wrote a bar the market had already mostly cleared and then the market kept going.

The number that decided the record, and the one we published in advance

The genuinely useful thing Z1’s window produced was not the pass. It was the boundary price.

On Sunday morning we published $76,744.29 as the exact settled close that would flip the answer to “best week since” from February 2024 to March 2023, and noted we were $572 short of it at press time. The week settled at $77,734.00 — $989.71 above the line. Sunday’s last seventeen hours moved the record from a six-month one to a two-and-a-half-year one.

The methodological point is the one we adopted as a rule yesterday and are now applying to our own output: a superlative computed on an open window is a live quantity with an unstated expiry. We printed the settlement time and the threshold price alongside Saturday’s and Sunday’s claims. Both of those claims are now retired. The only reason a reader could follow the change rather than being confused by it is that the boundary was published before it resolved.

Z2 — still failing, in our favour, and now at a series low

The marker, as written on 21 August: Binance BTCUSDT open interest, measured in coins, will reach 111,988 BTC or more on any daily reading on or before 28 August. Deadline type: arbitrary, sized to Jackson Hole.

Status: failing, and the gap widened. Monday’s daily reading is 105,531.28 BTC — the lowest in the series we have been tracking and 5.77% below the bar. Four days remain.

Z2 exists to falsify our own thesis. We have argued for six runs that this rally was not financed by leverage. If coin open interest had climbed back to its pre-rally level, that argument would be dead and the marker would say so in public. Instead it has gone the other way, and the reading that would have embarrassed us has not arrived.

Read that final row twice. The coin series is down 5.77%; the dollar series is up 16.21%. It is the same position. Every headline that said “open interest surged during bitcoin’s best week since 2023” quoted the dollar number, which price inflated for them. Read open interest in coins or not at all.

The funding floor: sixteen settlements, eight of them pinned exactly to baseline

The second leg of the same argument got stronger over the weekend and is now, in our view, the single most under-reported fact about this rally.

Sixteen consecutive eight-hourly settlements across a $14,834 rally and a $3,954 give-back, and the rate never once exceeded the 0.0100% baseline. The last eight consecutive settlements printed exactly 0.0100% — the floor, to four decimal places, for two and two-thirds days straight.

The interpretation matters and we will give both, as we have from the start. One reading: with no crowded long financing this move, there is no leverage to flush, and a pullback has to come from spot sellers rather than from a liquidation cascade. The competing reading: an unlevered rally has no marginal buyer waiting in the wings either, so the fuel is already spent. We do not know which is right. We know that both are more defensible than the reading offered by anyone quoting dollar open interest.

Benjamin Cowen, Bitcoin: It’s Almost Crunch Time. Embed verified against the YouTube oEmbed endpoint on 24 August 2026.

The other ratio: the size accounts are now more than two-to-one long

Binance publishes two long/short statistics that are routinely quoted as if they were the same number. They are currently telling opposite stories, and the divergence has widened for a fifth straight day.

The account ratio counts accounts: it says the retail population entered the week 2.2 to 1 long, capitulated into the rally, and has sat at roughly parity for five days. The position ratio weights by size among the largest accounts: it has risen for five consecutive days to 2.0859, a series high.

Two of those five increases were recorded during the 4.19% drawdown from Friday’s high. The size accounts bought the dip; the crowd neither bought it nor sold it. This kills the tidiest available bear story — that a retail squeeze is now unwinding as retail leaves. Retail already left, at parity, and it did so before the high.

A caution against ourselves: a rising top-trader position ratio into a series high is, historically, not a comfortable configuration. It means concentrated size is one-directional. That is exactly the setup that a single adverse catalyst — Wednesday’s core PCE, Friday’s Warsh keynote — can unwind quickly, funding floor or no funding floor.

Z3 — the window opens today, with zero rows in it

The marker, as written on 21 August: the sum of daily net spot bitcoin ETF flows for 24–27 August will be at or above $0, on completed rows only. Deadline type: scheduled.

Status: window open, no data. Farside’s last completed row is Friday 21 August. Monday’s flows publish on Tuesday. The relevant precondition is the one we surfaced on Sunday: over the record week, non-IBIT daily net went $137.3m / $45.7m / $232.5m / $103.3m / $68.2m. From Wednesday’s peak to Friday, the non-IBIT complex faded 70.7% against IBIT’s 15.9%. Friday’s IBIT share was 77.8%, the week’s highest concentration on the week’s smallest total. The question Z3 actually asks is not whether flows are positive but whether anyone other than BlackRock shows up.

The rest of the board

X2 required a Brent settle at or above $95.00 and was 0.65% away on Friday. Monday morning it is 2.05% away and moving in the wrong direction, into a sanctions announcement that would conventionally be read as bullish crude. That is either the market disbelieving the sanctions or the market having already priced them; the settle at 19:30 GMT will say more than the intraday tick.

Y1 required a 30-year close at or above 5.35% by 28 August. The 2026 high is 5.34%, set on 18 August, so the marker requires a new high for the year within four sessions — against explicit Treasury policy pushing the long end the other way. It sat at 5.238% at 06:10 UTC, 11.2bp away and falling. We flagged last week that describing this as “about 7bp away” was arithmetically true and analytically lazy. It is now 11.2bp away and the laziness has been corrected in the direction of it looking harder, not easier.

Difficulty, under the new rule

The 23 August retarget settled at −1.3122%, taking network difficulty from 127.48T to 125.807T. The next retarget, at block 965,664, is currently projecting −3.60% at 8.53% epoch progress, with 1,844 blocks remaining.

Under the rule this desk adopted yesterday, that projection is not a forecast. Our seven-snapshot post-mortem on the last epoch showed the estimate was wrong in sign for two consecutive days at mid-epoch and only converged inside 11 basis points at roughly 200 blocks out. We keep publishing the series because the volatility in the estimate is itself informative about hashrate; we no longer let it carry an argument before the epoch is more than 90% complete. Instantaneous hashrate is 887.1 EH/s.

Benjamin Cowen, Bitcoin: The Four Year Cycle Strikes Again — included as a counterweight to the constructive read above. Embed verified against the YouTube oEmbed endpoint on 24 August 2026.

Three new markers

Each is date-anchored, states its deadline type, and names the instrument it is written against.

A2 deserves one note on instrument, per the rule we wrote yesterday. The $71,440.63 bar is derived from the Binance monthly open of $62,887.88 and August 2021’s +13.60%. It is written to two decimal places against a venue whose Sunday close differed from Kraken’s by $3.50 and Coinbase’s by $4.64. At a 7.4% cushion that spread is immaterial; if the market comes within a percent of the bar, it will not be, and we will say so then.

What we are watching between now and Friday

Bessent at 2:00pm ET, and specifically whether secondary sanctions name Chinese financial institutions. Strategy’s 8-K, which settles A1 the moment it hits EDGAR. Tuesday’s ETF row, which is the first flow data since the high and the first datum in the Z3 window. Wednesday’s core PCE at 8:30am ET, which settles A3. And Friday’s Warsh keynote at 10:00am ET, the first of his chairmanship, at a symposium themed on payments and financial innovation — where X2 and Y1 both expire the same week.