Markers Monday: T1 Fails at $62,900, and the Hashrate Refuses to Agree With the Price

This column writes its calls down before the events and grades them after, in public, whichever way they land. Today: one clean failure, one measurement that cuts against the tape, and three new commitments — each anchored to a calendar date rather than to an event happening as scheduled, the fix we adopted after Friday’s cancelled SEC meeting produced this board’s first VOID grades.

T1: FAIL, on both sources

On Sunday we wrote T1 down as follows: the weekly candle closes at or above $63,220 at 00:00 UTC Monday. It closed at $62,900.00 on Binance’s BTCUSDT weekly kline, settled 23:59:59 UTC on August 16 (our own exchange pull), and at approximately $63,057 on CoinGecko’s aggregate four-hour candle into the same boundary. T1 FAILS — and it fails on both sources, by $320 and $163 respectively, which is the version of a grade we actually want. Markers that would flip depending on which price feed you asked are badly written; this one did not.

The context makes it worse rather than better. Reading the weekly series back, $62,900 is the lowest weekly close since the week beginning June 22, when bitcoin settled at $59,577. The four weekly closes before it: $63,570, $64,902, and before those $65,400 and $64,723. Bitcoin has now closed lower on the week in three of the last four weeks, and the weekly low of $62,535 undercut the prior week’s $62,300 only narrowly — a market grinding down, not breaking down.

Source: Binance BTCUSDT weekly klines, own pull 06:15 UTC August 17, 2026.

And then, having failed the test, the market bid. Binance’s Monday candle opened at $62,900 and had traded to $63,616 by the 06:00 UTC hourly close, up 1.14%, with the session low at $62,751 holding well above Friday’s $62,535. This is the pattern that punishes people who trade weekly closes mechanically: the level failed, and the immediate reaction was the opposite of what the failure implied.

The number that disagrees with the price: a 1.35-point swing in 24 hours

Here is the morning’s genuinely new observation, and it comes from a source nobody else is quoting because it requires pulling the data yourself. At 06:10 UTC on Sunday, our own mempool.space pull put the projected next difficulty adjustment at −1.63% with 52.2% of the period elapsed. At 06:10 UTC on Monday, the same endpoint put it at −0.28%, with 60.07% elapsed and 805 blocks remaining to the retarget at height 963,648. That is a 1.35-percentage-point swing in a single day. A projected retarget only moves that fast in one direction for one reason: blocks arrived materially faster than the ten-minute target, which means hashrate came onto the network.

The hashrate series says the same thing more directly. The same endpoint’s daily average printed 1,025 EH/s for the most recent period — the first reading above one zettahash per second to appear on this board — against 918 EH/s and 942 EH/s in the two preceding prints, with the instantaneous estimate at 906 EH/s. We attach the usual caution: short-window hashrate figures are inferred from block timing and are noisy, which is exactly why the instantaneous and daily-average numbers differ by more than 10%. But the difficulty projection is an independent confirmation of the same fact, computed a different way, and it moved hard.

Why it matters: for three weeks the bearish read on this market has been that price weakness plus flat flows equals a miner squeeze, with the negative retarget as the tell. A negative retarget was the forecast; it is now barely negative and trending toward flat. Miners are, on the evidence of the blocks themselves, adding machines into a $63,000 price and a week of ETF outflows. Either they know something the flow data does not, or the industry’s pivot into AI and high-performance compute has decoupled its capital expenditure from the bitcoin price — which is precisely the thesis our miner coverage has been building since MARA’s $611 million quarter. We have written a marker to force the question.

One more on-chain note from the same pull: bitcoin’s tip height was 962,843 at press time, and zero of the last fifteen blocks signalled bit 4 for BIP-110. With 805 blocks to the window’s end, the soft fork’s activation remains arithmetically out of reach, exactly as our August 8 coverage concluded. That is a call that has aged well and requires no further defence.

The whale divergence, one week on

The setup we framed on Sunday has not resolved. Wallets holding at least 10,000 BTC stood at 90, a six-month high, up six over eight weeks, with the 10-to-10,000 BTC band adding roughly $1.5 billion since July 29, per Santiment data reported by BeInCrypto. Against that, a single trader carried a 1,900 BTC short worth about $125 million at an average entry of $63,582, per crypto.news. Monday’s $63,616 print is the first time since that position was reported that spot has traded through the short’s average entry. That is not a squeeze; it is the first inch of one. The weekly close was supposed to arbitrate this and did not — it graded the price, not the positioning.

Board status and three new markers

U1 tests tactic versus policy. Two consecutive weekly filings disclosing bitcoin sales — 1,638 coins on August 3, 1,690 on August 10, both funding preferred dividends and STRC repurchases under the BTC Monetization Program adopted June 29, which authorises up to $1.25 billion of tactical sales — can still be read as opportunistic. A third makes it the operating model. One complication we will hold ourselves to when grading: the August 10 filing also disclosed 6,585,682 common shares sold for $653.1 million net. Strategy is not a company that has run out of ways to raise money; it is a company choosing to raise some of it by selling coins. We anchor to a filing dated on or before Friday, August 21 rather than to “today’s 8-K,” because filings slip and markers should not — the lesson the VOID grades taught us ten days ago.

U2 is the direct test of this morning’s hashrate read. If the settled retarget prints flat or positive, the miner-capitulation thesis is dead for this cycle and the AI-capex decoupling story owns the mining sector. If it prints meaningfully negative after all, the projection swing was luck-driven noise and we will say so. Note the bar is set at zero, away from both the current −0.28% projection and Sunday’s −1.63% — a marker that would be satisfied by doing nothing is not a marker.

U3 is deliberately uncomfortable. Bitcoin needs about 2.2% from Monday’s $63,616 to reach $65,000, and it has not closed a day above that level since August 3. Setting the bar on the day of Warsh’s Jackson Hole keynote makes it a joint test of the rate path and the bid — and it is the marker most likely to embarrass us, which is the point.

What would change our mind

The hashrate read is the fragile one. Difficulty projections drift as a period matures, and 60% elapsed is not 95% elapsed; a slow final 805 blocks would pull the estimate back toward Sunday’s figure and make this morning’s observation a footnote about luck. Equally, a fourth and fifth consecutive day of ETF outflows this week would restore the simple bearish reading: distribution into a weak close, with the mining data a lagging curiosity. And if Wednesday’s FOMC minutes read more hawkishly than the three-dissent framing already implies, the September odds spread — roughly 29% on Kalshi against as much as 82% in fed funds futures, a dispersion we print rather than average — resolves upward, and $65,000 by August 28 stops being uncomfortable and becomes unlikely.

Did bitcoin fail its weekly close test?

Yes. The August 10–16 weekly candle closed at $62,900 on Binance and roughly $63,057 on CoinGecko’s aggregate, both below the $63,220 bar. It is the lowest weekly close since the week of June 22.

Why did the bitcoin difficulty projection change so much in one day?

The projected retarget moved from −1.63% to −0.28% between Sunday and Monday morning because blocks arrived faster than the ten-minute target, which happens when hashrate joins the network. The retarget settles at block 963,648, roughly 805 blocks away as of August 17.

What is a zettahash?

One zettahash per second is 1,000 exahashes per second, or 10^21 hashes. mempool.space’s daily average estimate printed 1,025 EH/s in the latest period. Short-window hashrate figures are inferred from block timing and are noisy, so treat single prints with caution.

Is BIP-110 still able to activate?

No. Zero of the last fifteen blocks signalled bit 4 at press time, and only about 805 blocks remain in the signalling window against a 55% threshold. Activation in this window is arithmetically out of reach.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Cryptocurrencies 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.