Bitcoin 2026-08-02 daily brief — AI stance: bearish

Perpetual funding is neutral and should not be read as evidence for either side. The open-interest-weighted rate across exchanges — the right figure for any market-wide claim — annualizes to 9.37%, which sits below the roughly 11% per year that the exchange-default rate represents. That is at or under baseline: nobody is paying up to be long. The largest single venue runs cooler still at 7.12% annualized, 2.25 percentage points beneath the cross-exchange aggregate, which locates what modest long-carry demand exists offshore rather than at that venue. This is a meaningfully cleaner picture than the crowded-long conditions flagged in early June and mid-May, and it is one of the better arguments against a violent downside resolution: there is very little leveraged length stacked up to be forced out.

Positioning is light and shrinking, which is why this is a medium-confidence call rather than a high-confidence one. Market-wide futures open interest across all venues stands at $48.19B, down 5.6% from $51.06B on 07-21 while price fell 4.36% over the same span — open interest declining roughly in line with price is orderly deleveraging, not a positioning build in either direction. Options open interest at $24.76B looks like a collapse from $35.99B on 07-31, but that is the July monthly expiry rolling off the board, not a de-risking event; the practical consequence is that a large block of dealer gamma that had been pinning price into month-end is now gone, which frees the market to trend. Combined with 30-day implied volatility at the 3.3rd percentile of its trailing range, the setup is one where a range break is cheap to express and not currently priced.

The flush already happened and it was on the long side. The 07-31 session, a -3.0% day, took out $71.29M of longs. Since then liquidations have gone quiet — 08-02 saw $21.76M of shorts against $5.33M of longs, a 4.08-to-1 ratio, on a +0.94% bounce off 62,780, and total liquidation volume sits at just the 17.3rd percentile of the trailing year. Read together: the leveraged longs who were going to be shaken out already were, and the marginal aggressor over the last session was a short getting squeezed on a minor bounce. That argues any further downside grinds rather than gaps.

This is the most informative single input in the brief. The US-versus-offshore spot spread reads -9.46bp, just inside the ±10bp practical extreme, and the trend matters more than the snapshot: it has been negative for all thirteen sessions in the window and has deepened materially, averaging -7.5bp across 07-21 to 07-26 and -10.2bp across 07-27 to 08-02, with three prints beyond -10bp including -13.78bp on 07-30. Persistent negative readings mean offshore is leading and US-side buying is absent. This corroborates the record-low July ETF month through a fully independent channel, which matters here because the internal ETF flow series is stale and sign-inconsistent with the published figure — the spot spread is the trustworthy witness.

CONFLICT

Three things, in order of weight. First, US spot ETF demand has gone flat: July net inflows of roughly $205M were the smallest monthly total on record since launch. That is still positive — demand exhaustion, not distribution — but it removes the marginal bid that carried prior legs. The internal flow series is unusable here: it last prints 07-31, is null now, and disagrees with the published figure on sign, so this read rests on the published number alone. Second, miner economics have broken down structurally rather than cyclically: difficulty is down 19.1% from record highs and below year-earlier levels for only the second time in Bitcoin's history, with capital rotating to AI compute. The hash ribbons ratio at 0.9884 registers as a bottom trigger, but a difficulty decline driven by permanent capital reallocation is not the classic capitulate-then-recover pattern that signal was built on. Third, a live supply and confidence shock: the Coldcard seed-generation exploit has taken 1,367 BTC, roughly $89M, from 4,585 addresses across three waves, and pushed sub-1 BTC transfers to 39,600 BTC on Friday, the highest daily level since 2022. The 07-31 session carries the fingerprints — fees spiked to 938.75 sats per transaction against a roughly 450 norm across the surrounding two weeks, and address outputs hit 731,006 against a roughly 570k norm. Coins moving from self-custody to exchanges is potential sell supply, and it also means the on-chain readings for 07-31 and 08-01, including that 0.917 SOPR print, are partly migration artifact rather than genuine holder capitulation. On regulation, the SEC has frozen Nasdaq's QBTC options approval pending review through August 24, and the CLARITY Act has no scheduled floor vote with the Senate entering recess in about a week — no catalyst from either direction.

Hawkish. Fed funds sits at 3.63% with the 10-year at 4.68%, a 105bp gap that says the market is not pricing the policy rate lower. The FOMC held on July 28-29 while explicitly preserving the option to tighten further if inflation proves persistent, and the reaction confirmed the read: the 30-year rose more than 9bp to 5.193% and the 10-year 5bp to 4.657% while the 2-year slipped 4bp to 4.236% — a bearish steepening driven by term premium, not by growth optimism. The one offsetting element is M2 growing 5.53% year over year, which is not a restrictive liquidity backdrop. On sentiment, Fear & Greed reads 25, Extreme Fear, but that print is dated 07-18 and is fifteen days stale as of this data date — it should be treated as a contrarian input of decayed reliability, not a live gauge. Nonfarm payrolls on 08-07 is the binding event inside this horizon; CPI on 08-12 sits just outside it.

The 10-year at 4.68% has been grinding higher within a tight band — 4.61% on 07-29, 4.71% on 07-24 — with no directional resolution, but the long end is where the move is: the 30-year at 5.193% after the July FOMC. The Treasury Quarterly Refunding Announcement on 08-05 lands squarely on that pressure point, since it sets coupon auction sizes and therefore long-end duration supply. There is no credit-spread feed in this data, so this brief makes no claim about credit conditions.

DXY at 99.71, having broken below 100 from 101.50 on 07-27 — a 1.76% slide in six sessions. A falling dollar is normally a direct tailwind for BTC. It has not been one here: over the same 07-21 to 08-02 window in which DXY fell 1.45% from 101.18, BTC fell 4.36% from 66,257. That non-response is the single most important macro fact in this brief.

Risk-on, and BTC is not participating. The S&P closed 7,489.72 on 07-31, up 2.37% from 7,316.15 on 07-29, while VIX collapsed from 20.66 to 15.99 over the same two sessions. BTC fell 1.31% on that same 07-31 session. Note both series last print on 07-31, three days before this data date, so the equity read is carried forward rather than current.

With 30-day realized volatility at 31.03%, a one-sigma seven-day move is about 4.3%, putting the band at roughly 60,650 to 66,100 — which barely escapes the thirty-day range at either end. That is the honest calibration: the base case is lower within the range, not a breakdown. On the way down, the first test is the fourteen-day floor at 62,780 from 08-01, then the thirty-day low at 61,849, about 2.4% below spot and untested since before this window opened. Below that, the $60,000 handle is where the most crowded open position on Deribit sits, and dealer hedging around a heavily-owned put strike can accelerate an approach to it. The 60-day low at 58,519 is 7.7% down and is not a reasonable seven-day target absent a macro shock. What argues against a violent move is the leverage picture: cross-exchange futures open interest has already bled to $48.19B, funding is below neutral, and total liquidations sit at the 17.3rd percentile — the 07-31 session, which cost longs $71.29M on a -3.0% day, likely already cleared the fuel. The asymmetry worth naming is that implied volatility at the 3.3rd percentile means the market has priced almost no probability of the range breaking at all, into a high-impact payrolls print on 08-07 and with the July expiry's pinning gamma now gone.

low

One data conflict, flagged and set aside: the internal 25-delta skew reads +0.069 at the 83rd percentile and is labelled call-premium, which contradicts the $60,000 put being the most-owned line on Deribit. That field ships with an explicit disclaimer that it is operational context and not a quantitative gate, so this brief leans on the observable — the crowded put and the 3.3rd-percentile volatility — and treats the skew label as unresolved.