Everything Bitcoin wanted showed up. Bitcoin didn't.

This week the market handed Bitcoin every single thing it has been asking for.

Inflation cooled. July CPI printed 3.4%, down from 3.5%, with core at 2.5%. The rate hike that has been sitting on every crypto chart since the spring quietly stopped being the base case, with CME FedWatch putting September odds at 67% on July 31 and 32% by Thursday. Money got braver. The S&P set a record close on Thursday, small caps ran to a record of their own, and volatility fell to its lowest since December. Then Friday piled on two more soft prints, with July retail sales at minus 0.6% against an expected plus 0.1% and consumer sentiment at 51.0 against an expected 54.5.

Five sessions. The entire wish list, and then some.

The S&P booked a third straight winning week. At Friday's publication cut, Bitcoin was $62,891, down 3.1% on the week, having spent the afternoon underneath $63,000.

That is the issue. Not "is the Fed done," which is what everybody else spent the week on. The question worth your Friday is why the asset that is supposed to be the purest expression of easier money did not move, when money got easier.

The answer has three parts: If you have been following the show, we stress-tested the $100 trillion thesis on Tuesday and sat with Eric Krown last week on whether the low is already in. This is the written case underneath both, with the flow data the videos did not have and the level that actually decides it.

One piece of housekeeping first, because a lot of people spent today expecting something that never happened.

The SEC was scheduled to meet at 10:00 this morning, Eastern, to vote on proposing "Regulation Crypto": a startup exemption, a larger fundraising exemption, and a safe harbor for tokens whose founding team has stepped back. After CLARITY stalled in the Senate it became the industry's main hope of getting rules in 2026 at all.

It was cancelled the night before. The notice is one sentence long: "The Open Meeting scheduled for Friday, August 14, 2026, at 10:00 a.m. has been cancelled." No reason, no new date. An SEC spokesperson told CoinDesk it was an unforeseen scheduling issue and that the meeting will move.

Take them at their word on intent. A rulemaking this far along does not die on a Thursday afternoon. But two things make it more than a diary clash. In the same 24 hours, CoinDesk reported that the agency had again delayed its separate innovation exemption covering tokenized securities, and as of tonight's close neither effort had a new public date. Chair Paul Atkins has said nothing publicly. He keynotes in Jackson Hole on Tuesday.

So the one piece of US crypto rulemaking with momentum went quiet in the same month CLARITY slipped to a scheduled September 15 cloture vote. Whatever the reason, token issuers end this week with no new tailored legal path to raise and no date on the calendar that changes it.

The inflation scare has an expiry date, and you can read it off their own table

Start with a number that is ours. Truflation's live US inflation index, the daily feed we pay for, reads 2.20%. The official BLS figure is 3.40%. A gap of 1.2 points.

Be careful with that gap, because I have seen people misuse it. Those two indexes do not measure the same basket the same way, and one is not simply a fresher version of the other. So do not read the difference as an error the BLS is about to admit to. Read it as a question: what is inside the official number that is not inside ours?

It is not a one-off, either. The gap shows up in all three pairs Truflation publishes against an official statistic, and it points the same way every time.

They answer the question themselves, in a table nobody opens. Of July's 3.4% annual increase, energy supplied 0.982 of a percentage point. Call it 29% of the headline, from a category that is about 7% of what a household actually buys.

And that energy number is not about this month. Energy prices fell 1.5% in July and gasoline fell 2.9%. Almost all of the plus 14.7% traces to a single three-month step between February and May, when the seasonally adjusted energy index jumped 19.6% on the Hormuz shock. Since May it has come back down 7.1%.

So the official inflation rate is carrying a price shock that already happened and has already partly reversed. Not an opinion. Arithmetic.

Here is the part I have not seen anyone write down. Freeze the not-seasonally-adjusted index, the one used for the published annual rate, exactly where it sits today. No decline, no rebound, nothing moves. Purely because of when the shock hit the comparison window, energy's annual rate would still read about plus 17.3% in February 2027, then plus 4.8% in March and minus 1.4% in April.

That is roughly nine-tenths of a point coming off the official inflation rate between February and March, and another four-tenths by April, and it requires nothing to happen. No policy. No demand destruction. No price falling anywhere. Just the comparison base rolling from February to March to April 2026.

There is a specific two-month window when this scare ends, and it is already on the calendar.

Now the obvious objection, and it got louder today. That arithmetic holds the energy index still. Oil does not have to cooperate. Brent finished the week at $88.52, up about 6%, after Treasury Secretary Scott Bessent said the US can maintain a naval blockade of Iran indefinitely and promised "measures like have never been seen in the history of economic isolation of a country" next week. A second Hormuz shock would reload the base and push the whole schedule out. That is the live risk to this call, and it is the one to watch rather than the Fed.

Now the other side, because I am not here to sell you a clean story. Thursday's producer prices were unchanged on the month, which is what got the cheer, but over twelve months final demand still ran 4.7%, and the core measure stripping out food, energy and trade services rose 0.4% on the month and 4.7% on the year. Headline goods fell along with food and energy while the core went the other way. Portfolio management fees alone rose 6.5%, and those feed straight into core PCE on August 26.

The pipeline is still warm while the consumer headline cools for a reason with an expiry date. Both are true. If you only heard one of them this week, you heard the one the tape wanted.

So who actually sold?

If the macro broke crypto's way, and crypto fell anyway, somebody was selling into it. We can name part of that.

Messari's July report caught it first: the month's improvement "was driven more by an easing of selling pressure than by renewed accumulation." Less bad, not good.

Then it got concrete. Corporate treasury companies turned net sellers of Bitcoin for the first time, about $140 million of net disposals in July. Set that against what those same buyers did in April, which was $4.1 billion of net purchases. That is not a bid getting scared. That is a bid inverting.

And we know a name. Strategy disclosed on August 10 that it sold 1,690 BTC between August 3 and 9 at an average of $64,262, cutting its holdings to 840,447 coins against an average cost of $75,385.

Then read the next line of the filing, which is the part nobody quoted. Those coins raised $108.6 million. Strategy used $108.6 million to buy back 1,152,020 shares of its own preferred stock. On the filing's own rounded figures, it sold just enough Bitcoin to cover the buyback.

Sit with that. The loudest structural buyer this market has ever had sold coins more than $11,000 below its own average cost, and routed the proceeds straight into its capital structure. That is not a view on Bitcoin. It is a balance sheet with obligations.

And on Friday those obligations got heavier. MSCI opened a consultation on screening non-operating companies out of its global indexes, and its own May simulation deleted Strategy and Metaplanet, among three names, from ACWI IMI. Strategy has $23.9 billion of free-float-adjusted market cap riding on it, feedback closes September 30, and the decision lands October 16. The company's response: "Index providers should measure markets, not decide which assets companies are allowed to own." Its stock fell about 7% on the week. If that screen passes, the forced seller is not a fund manager with a view. It is an index.

The ETFs did it in miniature, and the pressure lasted through Friday. Spot Bitcoin funds went out $144.6 million Monday, took in $7.8 million Tuesday, then went out $61.1 million Wednesday, $131.1 million Thursday and $56.2 million Friday. Net of $385.2 million over five days.

One detail inside Thursday's number looked worth more than the number. The selling was ARK at minus $58.8 million and Fidelity at minus $55.1 million. BlackRock's IBIT, the biggest fund in the category, was only minus $5.7 million. Then Friday took that distinction away. IBIT led the day's selling at minus $55.5 million.

None of that proves it is the whole story. Markets are not a single-cause machine and I cannot hand you a clean attribution for one week's return. What I can hand you is the part with receipts on it: while the macro improved, identifiable supply was hitting the tape, and some of it came from balance-sheet pressure rather than a fresh view on Bitcoin. That is a very different problem from nobody wanting this.

Selling that comes from balance-sheet pressure runs out. Selling that comes from disbelief does not.

What that leaves

On July 23, Messari published a study of high timeframe conditions. One of its signals takes the ratio of the Nasdaq 100 to Bitcoin, runs a 14-week RSI on it, then smooths that with a 14-week average. It is not a trading indicator. It turns over on the scale of multi-year cycles.

On that date it read 72.6, the highest in the sixteen years of its history, 4.1 points above the previous record set in September 2022. That measure has spent 0.35% of its life above 70. It has done it four times, ever: February 2015, February 2019, August 2022, and the stretch that began in late January this year. As of the report, the current episode had run 24 weeks, already the longest of the four.

Inverted, that reading says Bitcoin was more oversold against the Nasdaq, on a high timeframe, than at any observation on record. And the broad move since, with stocks setting records while Bitcoin fell, has pushed the underlying ratio further that way.

Two more from the same report, dated for the same reason. Realized price, the aggregate on-chain cost basis of every coin in circulation, sat near $53,000, and only 12% of Bitcoin's price history has been spent below that line. On cycle structure, the 2013, 2017 and 2021 bear markets each set their price low by week 60 from the all-time high. Messari called this cycle week 40 on July 23.

Now every caveat, because a signal delivered without its failure modes is a sales pitch. Realized price is not a floor: Messari is explicit that a drop to or below it would be normal, since every historical bear low traded at a discount to it. Short horizons are noise, and the 2022 cohort of this signal was down 29.1% at 120 days before resolving to plus 397% at three years. The returns decay, with each cycle's three-year figure coming in at roughly a quarter to a third of the one before. And Messari's own warning, which I would rather quote than bury: these signals "should not be treated as independent corroboration," because their simultaneous presence "is closer to one observation measured several ways rather than several unique and independent observations." All three are, underneath, the same fact wearing three outfits. Bitcoin has fallen a long way for a long time.

The scoreboard this week, kept honest across every lane. The numbers are the point, so here they are in one place.

The level that decides it

Bitcoin printed a shade over $65,300 on Monday and spent the rest of the week giving it back, reading $62,891 at Friday's publication cut. It has not managed a daily close above $65,000 since July 27, and was about a hundred dollars above $62,800, which is where the weekly cloud support sits.

On Tuesday's show I disagreed with Jordi about which chart to watch, and I want that disagreement in writing, because it is the most useful thing I can give you this week.

Jordi's line is the 200-week moving average, now around $63,900. Bitcoin reclaimed it early in the week and gave it back by Friday. By his framework, the signal that looked live on Tuesday was no longer live at Friday's publication cut.

I do not use a single moving average for a regime question. One line can tell you a new story four times a month. What I want is the weekly Compass, where price and the money line need to flip together out of a bear regime, and that line is sitting at roughly $80,000.

We are at $62,891. So the honest position, and the one I am actually in: the setup is the most interesting it has been in this entire drawdown, and my own trigger has not fired. Those two are not in conflict. A setup is a condition. A trigger is a decision. Confusing them is how people spend a bear market slowly being right and quickly being broke.

If it flips, you will hear it from me. It has not flipped.

Protect your coins, and please read this one properly rather than skimming it. The ColdCard seed-generation flaw is worse than any headline number, because the failure sat at the moment the key was created. Affected devices drew their randomness from a far weaker source than intended, so an attacker can reconstruct the key without ever touching your hardware. Air-gapping bought you nothing. Galaxy Research now has 1,778.84 coins, about $112.7 million, confirmed stolen, with suspected but unconfirmed footprints that would take the total to 2,417.35 BTC, about $153 million.

Are you exposed? The affected set is Mk2 and Mk3 on firmware 4.0.1 through 4.1.9, Mk4 and Mk5 before 5.6.0 standard or 6.6.0X Edge, and Q before 1.5.0Q standard or 6.6.0QX Edge. Standard and Edge are separate release tracks, so a higher Edge version number does not mean you are patched. There is one real exemption: if you generated your seed with at least 50 fair, independent dice rolls and that sequence was never recorded or exposed, Coinkite does not consider the seed at risk from this issue alone.

A passphrase is not an exemption, and I had this wrong until I reread the advisory. A strong, unique BIP-39 passphrase does add a real second barrier, because an attacker needs the passphrase as well as the weakened seed. But Coinkite is unambiguous that it does not fix the underlying problem: "A strong passphrase reduces the immediate exposure, but it does not repair the affected seed; passphrase users should also migrate as soon as practical." If your passphrase is short, reused, quoted or patterned, or you are simply unsure how strong it is, treat the funds as at risk and move now.

Firmware alone does not save you either. Updating does not repair a seed that was already generated by the bad code. You install the fixed firmware and then create an entirely new wallet.

Then migrate carefully. Verify the new backup before you deposit anything. Verify the receiving address on the device screen. Send one small test transaction and confirm it lands before you move the rest. Keep the old backup until the migration is complete. Take those steps from Coinkite's own advisory, not from any summary, including this one.

The thing that could reprice everything next week. Treasury Secretary Scott Bessent said Friday the US can maintain a naval blockade of Iran indefinitely, and told markets to "watch this space for more announcements coming next week because we are going to apply measures like have never been seen in the history of economic isolation of a country." Roughly a fifth of the world's daily oil and LNG moves through Hormuz. That is why Brent added 6% this week and why yields rose on a day of bad US data. Explicit guidance that more lands next week is rare. Take it at face value.

The calendar that matters. Next week, whatever Bessent announces on Iran. Aug 18, Atkins keynotes in Jackson Hole. Aug 26, core PCE at 8:30am Eastern and Nvidia's results around 4:20pm, the month's two biggest inputs on one date. Aug 27 to 29, the Fed's Jackson Hole symposium. Late August, the next scheduled HYPE unlock. Sept 15, the Senate's cloture vote on the CLARITY Act, which opens debate rather than passing anything and needs 60. Sept 15 and 16, the FOMC.

Scoring what this letter has said, and this week that starts with a miss.

The call that half broke. Issue 8 argued that rising sticky inflation plus a cracking labor market was a trap rather than an all clear. The labor half held: payrolls fell 23,000 with another 103,000 revised away, and claims have not confirmed a recovery. The inflation half turned against us. Our own live feed went from 2.32% to 2.20% and the official headline fell to 3.4%. What survives is narrower than what I claimed: core producer prices stayed warm at 0.4% on the month, and our wage index actually rose to 4.46%. The sticky side did not vanish. But "inflation is climbing" was the wrong tense, and it was our tense.

The hook that inverted. Last week's headline was "inflation went up while oil went down." This week our oil index climbed about 11% and Brent added 6%, while our live inflation read went down. The exact mirror, seven days later. A letter that only reports the weeks it was right is not worth the time it takes to read.

Our own source missed too. Truflation's July CPI forecast was 3.60% against an official 3.40%.

And one from the show. On Tuesday I said $100 trillion would make crypto 6.4 times the size of gold, using $15.7 trillion for gold. If you looked it up you probably found gold called a $31 trillion market and wondered which of us could not use a calculator. Both are right, and that is the useful part. The $15.7 trillion is investable gold, from the Goldman Sachs study the whole comparison came from, which strips out jewellery. The $31 trillion is every ounce ever mined, wedding rings included. Same metal, two honest answers, and the ratio swings from 6.4x to 3.2x depending purely on which one you reach for. My argument on the show was that the denominator is the argument, which turns out to be more true than I made it sound. What I did get wrong is vintage: Goldman struck that table near $4,000 gold, and at today's $4,374 the honest multiple on the same measure is about 5.8 times, not 6.4.

The boring one keeps paying. Hyperliquid's circulating supply is still 222,445,714, unchanged through another cycle. Scheduled is not claimed. And when we went to date the next unlock, the trackers disagreed with each other about when it even is. The calendar is an estimate, the chain is the fact.

Still open. September's FOMC. Whether treasury company selling continues or was a one-quarter balance sheet event. And Messari's cycle window, which we will refresh rather than assume.

There were two ways to read this week and both of them were lazy. The first was to look at record equities and collapsing hike odds and call it permission. The second was to look at a red crypto week and call the thesis broken.

The evidence supports something less satisfying than either. The macro headwind genuinely is easing, on a schedule we can read off a government table, though Bessent may yet reload it. Identifiable supply genuinely was hitting the tape at the same time, on four of five days, and some of it came from balance-sheet pressure rather than a market view. By Friday, an index provider had put Strategy's membership on the table too. And the condition that combination has produced is, by at least one long-horizon measure, the most extreme in its recorded history.

That is what a bottom looks like from the inside. It is also what the middle of a longer grind looks like from the inside. They are indistinguishable in real time, which is precisely why I would rather hand you $80,000 on a weekly chart than a feeling.

Eric Krown said something last week that has stuck with me: bear markets end on bad news, not good news. Everybody waits for a green light. What actually happens is the news stays ugly and price stops caring. This week we got the opposite, good news and price that would not respond, which is not the same thing and is not the signal. Five sessions of it, ending lower than it started.

So the plan does not move. Stay liquid. Keep the quality list growing. Let the trigger fire before you size up. And be a little suspicious of any week where the story is this tidy, including this one.

See you next Friday.

Kyle

This letter is free, weekly, and complete on its own, because the read has to earn your time before anything under it earns a dollar. The layer beneath, the names, the sizing and the entries I am actually taking, lives in the Data Room. The one thing I will say out here: the difference between a setup and a trigger is the entire job, and the work of turning one into the other is what is inside.

No rush from me. Read a few more Fridays, then decide.