Bitcoin 2026-07-31 daily brief — AI stance: bearish

Open-interest-weighted funding across venues annualizes to 5.39% — roughly half the neutral baseline implied by the standard exchange default. Perpetual longs are paying less than the flat-market rate, which means there is no leverage crowding to unwind and no long positioning to squeeze. The single major venue in the feed prints 6.64% annualized, 1.25 percentage points above the all-venue weighted figure; that says whatever modest crowding exists is concentrated at that one venue rather than distributed across the market, and it should not be read as a spread between two market-wide measures. The path matters more than the level: the weighted rate had built from 0.0049% per 8 hours on July 19 to 0.0085% by July 28-30 — around 9.4% annualized, approaching but never reaching the neutral baseline — and then halved back to 0.0049% on July 31 as $71.3M of longs were liquidated. That is a leverage build getting flushed rather than leverage steadily draining, and it leaves the book cleaner today than at any point in the past week. This is a genuinely constructive detail sitting inside a bearish view, and it constrains that view: the downside case here has to be driven by spot supply, not by a liquidation cascade, which is why the expected move is a range grind rather than a break.

Market-wide futures open interest is $48.18B, down 5.6% from the July 21 peak of $51.06B and roughly flat across the 14-day window — no net leverage build into the decline, which is deleveraging rather than positioning for a move. Options open interest tells the opposite story: $35.99B, a 14-day high, up 13.5% from $31.70B on July 19, against options volume of $2.62B. Risk is rotating out of linear leverage and into optionality, and it is doing so while 30-day at-the-money implied volatility sits at the 3.3rd percentile of its 90-day range and the term structure is in contango with 90-day implied above 30-day. Convexity is being accumulated cheaply. What makes that a bearish detail rather than a neutral one is the skew: at the 23.3rd percentile of its 30-day range, downside protection specifically is not what is being bought. A market long cheap optionality but light on puts is a market that would have to chase hedges on a break of the range floor rather than being cushioned by dealer flow already in place.

$71.29M of longs against $15.34M of shorts on July 31, total activity at the 62.7th percentile of the past year — elevated but not extreme. This is the second $70M+ long flush in five sessions, following $71.68M on July 27. The mirror image sits at the other end of the range: short liquidations of $52.23M and $48.09M on July 20-21 as spot printed the 66,257 high. Both boundaries of the 61,439-66,257 range are now liquidation graveyards, which is what a two-month distribution of leveraged positioning inside a range produces. The asymmetry that matters is recency — the last two flushes were both long-side, with no offsetting squeeze, so the pain has been one-directional for a week.

The US-versus-offshore spot spread is -8.97 bps and has been negative on every one of the 13 sessions in the 14-day window, deepening from -4.22 bps on July 23 through -12.33 bps on July 26 to a low of -13.78 bps on July 30 — that low is close to the ±15 bps p99 extreme and well past the ±10 bps practical decile threshold. A persistently negative spread historically tags offshore-led, risk-off tape with the domestic institutional bid absent, and that is precisely coherent with a 30-day ETF flow average in the 8.8th percentile: the same absent US buyer shows up in both series. The single-session improvement from -13.78 to -8.97 bps is the first genuine crack in that fortnight-long pattern, and it lines up with the reported $233.1M creation. Treat it as one datapoint, not a trend reversal — the spread needs to cross zero, not merely become less negative, before the domestic-bid thesis changes.

CONFLICT

The dominant micro fact is a month of structural ETF redemption: the 30-day net flow average is -1,158 BTC/day, sitting at the 8.8th historical percentile, with excess absorption of -2,760 BTC and vehicles redeeming 4.62x daily miner issuance. Against that, the most recent reporting day was a $233.1M net creation led by IBIT at $183.4M, the strongest single day in over three weeks, turning weekly net flows positive at $203.84M. A data caveat that a careful reader should have: the local feed's largest recent creation, +3,597 BTC stamped July 30, converts to $233.1M at that day's close of 64,803 — the identical figure the press attributes to July 31 — while the local feed shows -2,269 BTC for July 31 itself. One of the two is misdated by a day. The 30-day structural picture is unaffected under either reading, so the thesis does not depend on resolving it, but no weight should be placed on the single most recent print. Miner economics are compressing: hash ribbons at 0.9891 (10.1st percentile) signal hashrate contraction and Puell at 0.797 (28th percentile) signals depressed revenue, with Riot reporting August 5 and MARA August 6 into that. On the chain itself, a Coldcard firmware RNG flaw let attackers sweep 594 BTC from around 500 wallets in 25 minutes on July 31 — and on that exact day transaction count FELL 14% (712,078 to 610,890) while distinct output addresses rose 23% to 731,006 and fees per transaction more than doubled from 446 to 939 sats. Fewer transactions, each fanning out to more addresses, paying up for blockspace, is a sweep-and-migration signature; it is consistent with an emergency wallet exodus and should not be read as organic demand. Regulatory optionality has been repriced out: the CLARITY Act has passed the House and a Senate committee but has no floor vote or cloture motion, with early August the practical cut-off before recess, and Polymarket cut 2026 passage odds to 28% on July 30 from a February peak of 82%.

Hawkish, and the direction of travel matters more than the level. Fed funds sits at 3.63% (a 3.50-3.75% band) after the July 29 FOMC held for a fifth consecutive meeting on a 9-3 vote in which all three dissents were in favour of a HIKE — not a cut. Futures subsequently priced roughly a 72% probability of a quarter-point hike in September. For a zero-yield asset that is a regime change in kind, not degree: the question has flipped from when policy eases to whether it tightens. Two partial offsets: M2 is still expanding at +5.53% year-over-year, so the liquidity backdrop has not turned outright restrictive, and the 10-year at 4.68% has risen only 11bp over the past fortnight. The live sentiment gauge reads 25 — Extreme Fear — but it is stamped July 18, thirteen days stale, and it sits against a VIX of 15.99 and an S&P at 7,489.72, so it is measuring crypto-specific despair rather than any macro stress.

The 10-year is 4.68%, up 11bp from 4.57% on July 19, with the 30-year having risen more than 9bp on the July 29 decision. Rising long yields alongside a falling dollar is a fiscal/term-premium signature rather than a growth signature, and the August 5 Treasury Quarterly Refunding Announcement is the near-term test of exactly that. There is no credit-spread feed in this dataset — that is a genuine blind spot in the read, not evidence that credit is calm, and it should be treated as an unmeasured risk rather than a benign one.

DXY 99.803, down from 101.502 on July 27 — a 1.67% four-session slide that broke the 100 handle. In a normal transmission a dollar breakdown of that size is a clean tailwind for BTC over days to weeks. It did not transmit: on the same four sessions spot went 63,750 to 62,888 and posted its worst day in a month. That non-response is the single most informative observation in this brief. Note also that the dollar fell while long yields rose (30-year +9bp on the Fed decision day) and gold trades at $4,098.6 — that combination is a term-premium/debasement configuration that gold is expressing and BTC is conspicuously not.

Risk-on, with an unresolved overhang. The S&P closed 7,489.72, recovering from 7,316.15 on July 29 to within 0.26% of the 14-day high of 7,509.20, while VIX fell from 20.66 on July 29 to 15.99. Against that, the July 27 tape showed equity positioning being cut on the scale of Big Tech AI capital expenditure, and Brent fell below $90 only after a pause in US airstrikes on Iran — a de-escalation that is a pause, not a settlement. So the equity signal is genuinely supportive at a one-week horizon but built on two conditions that can reverse quickly.

First support is the 30-day low at 61,439, only 2.3% below spot. With 7-day realized volatility at 34.36% annualized, a one-standard-deviation move over the horizon is about 4.8%, so 61,439 is roughly half a sigma away — a touch of it is more likely than not over the next week, call it around 60%. The June low at 58,519 is 6.95% below, about 1.46 sigma, which puts a touch closer to one-in-seven than to a base case. Two things fatten the left tail beyond what the volatility number implies: 25-delta skew at the 23.3rd percentile means the market is thinly hedged for exactly that break, and a Nonfarm Payrolls print lands August 7 into implied volatility sitting at the 6.3rd percentile of the year, so any repricing starts from a floor. Two things cap it. First, there is very little forced-selling fuel — open-interest-weighted funding at 5.39% annualized is half the neutral baseline and futures open interest has already bled from $51.06B to $48.18B, so a cascade below the range floor has to be manufactured by spot sellers rather than by liquidations. Second, valuation: reserve risk at the 2.1st percentile and MVRV-Z at the 21st mean anything under roughly 58,500 puts the market at levels that have historically attracted persistent accumulation. The realistic distribution is a grind toward 61,400 with a fat middle, not a trend break; a sustained close below 58,519 would require a catalyst not currently visible in the data.

low