Daily Report — 07/30/26 · “The Liquidation Ended, The Repricing Did Not Start”

Scorecard — Grading Wednesday’s Report

“The Capex Referendum” earns an A on its central mechanism, an A on its two structural warnings, and an F on its rotation list — and the F is the more useful grade, because it is the fourth consecutive cycle in which the same screen has produced names that do not perform.

What it got right. The report’s core claim was that the market would reward the company that cut spending and punish the company that raised it. Thursday night inverted that and the report was still right for the right reason, because the mechanism it identified — that this market prices the AI capital-spending chain and not the consumer franchise — is exactly what produced the result. Amazon lifted its 2026 cash capital budget to two hundred and twenty billion dollars, from about two hundred, and said explicitly that higher memory costs were the reason. The stock rose nine percent after hours. Apple beat earnings by thirteen cents, beat on revenue, grew iPhone twenty-two percent and Mac twenty-nine, and fell six percent on a services and China miss. The report also said that the way to own a memory bottom is the basket rather than the single names. That call is the single best thing produced in the last five cycles: the memory basket rose sixteen and seven-tenths percent and is now the only name in the entire file besides Korea with a genuine, gate-clearing institutional accumulation footprint.

What it got wrong, and it is not close. Wednesday’s report named four stocks that cleared all three of its rotation gates. Every one of them was red on the biggest up day of the month.

The framework has a rule for this and it applies now. A tell that fails three times is wrong, not early. The search giant is on its third failure and comes off the list entirely — it was promoted to a tier on Wednesday on the strength of the best options structure in that file, and it has now produced three consecutive sessions of nothing. The other three are on failure two and are suspended rather than killed. But the deeper lesson is about the screen itself, and it is incorporated into this report: the print-count gate is necessary and it is not sufficient. A name can print fifteen separate institutional crosses with a positive net and still be distribution, if the tape is doing that on shrinking volume. From today forward the screen requires a thick tape and expanding day-over-day volume, and that single addition removes three of Wednesday’s four names before they are ever written down.

The secondary list did better and deserves its credit. The memory basket rose sixteen and seven-tenths percent; the cooling name rose two and a half and held its accumulation tag; the beverage name was flat. Against them, the Indian IT name fell five and two-tenths percent and the printer-and-PC name fell five and a half. Two clear wins, one hold, two clear losses out of the eight that could be measured.

And the two structural warnings both paid. The report said the biggest chip name’s two-hundred line was a financing box rather than a floor. It closed Thursday at one hundred ninety-five and four cents against eighty-four thousand contracts of open interest sitting at the one-ninety-five strike — pinned to the penny by the mechanics the report described. And the report’s instruction on the AI-financing sleeve was absence, not shorts. That cohort rose seventeen to thirty-one percent on Thursday. Anyone who read “avoid” as “short” lost a great deal of money in one session; anyone who read it as written lost nothing.

---

The Read — A Block Sale That Never Printed

The proximate cause of Thursday was public before the opening bell and it is not a story about earnings, the Fed, or an improvement in anything. A levered research fund whose assets had fallen from about forty-five billion dollars to roughly ten, running something close to four times leverage into the semiconductor drawdown, sold the bulk of its public equity portfolio to a single large market-maker in a negotiated block. A second multi-strategy firm bid and lost. The names reported as the core of that book were the largest chip designer, its main competitor, the database company and the biggest memory manufacturer — which is to say, precisely the complex that had been bleeding for six straight sessions.

The mechanical point is the whole story and it is worth being slow about. A negotiated block transfer does not print on the public tape as selling. The shares move from one balance sheet to another at an agreed price, off-exchange, in one transaction. What the market experiences is not a seller arriving; it is a seller disappearing. Every day for six sessions there had been a forced, price-insensitive supplier of these shares. On Thursday morning there was not. And the shares that supplier had been feeding into the market had, over those six sessions, been absorbed by a short base now carrying implied volatilities between eighty-five and one hundred sixty-three, punitive borrow costs, and dealers positioned short gamma below the market — which means dealers are mechanically forced to buy as prices rise.

Remove the seller from that configuration and you do not get a rally. You get a vacuum. The neocloud miner rose thirty and a half percent. The fuel-cell company rose twenty-six and a half. The storage name rose twenty-six. The AI-infrastructure name rose twenty-seven. The Korean memory proxy rose seventeen and a half. And the largest, highest-quality, most widely-held chip designer in the world — the one every long-term investor would buy first at a genuine bottom — rose two and sixty-five hundredths percent.

Real bottoms are led by quality, because quality is what patient money buys first. Squeezes are led by whatever carries the largest short interest, because that is what has to be bought regardless of price. Thursday was the second kind, and the two-and-a-half-percent print in the best asset in the group is the single cleanest piece of evidence in this entire file.

Was Wednesday the Low? Yes. Is the Correction Over? No.

These are two different claims and the evidence splits cleanly between them. The report will take a view on each.

Wednesday almost certainly was a low, and the case is mechanical rather than hopeful. The marginal seller is gone and cannot come back — the position has already changed hands. The retracement math confirms how violent the absorption was: across the eleven worst-hit names in the complex, the median single-session retracement of a seventeen-to-thirty-seven percent decline was fifty-four percent. The fuel-cell name retraced seventy percent of its decline in one day. The neocloud miner retraced seventy-five. Four names in the AI-infrastructure complex are already trading above last Friday’s close. If the question is whether Wednesday’s low gets revisited quickly, the honest answer is probably not.

But this is not the bottom, and it fails on every gate this framework has. A bottom is an accumulation event. This was a supply-clearing event. Six independent places in the data say so:

One. The institutional share ladder finished dead even.

Across six hundred and eighteen names, the fifteen-day accumulation-versus-distribution pattern finished at one hundred twenty-three accumulating and one hundred twenty-three distributing, with thirty in conflict and three hundred forty-two showing no pattern at all. On a day the Nasdaq rose three and a third percent. A bottom does not produce a coin flip.

Two. The seventeen names being accumulated hardest contain no technology at all.

Only seventeen names in six hundred and eighteen carry a strong accumulation tag. They are: the travel platform, the insurer, the specialty-metals maker, the Michigan utility, the warehouse retailer, the solar manufacturer, the exchange operator, the packaging company, the life insurer, the eastern railroad, the fertiliser company, the steelmaker, the Pittsburgh bank, the mall REIT, the telehealth company, the Japanese automaker, and the pharmaceutical-packaging company. Not one semiconductor. Not one memory name. Not one neocloud. Not one megacap. The list is insurers, rails, utilities, steel, fertiliser and a REIT. That is what institutional money was actually buying while the screen was green.

Three. Three of the thirteen most-distributed names were among the day’s biggest winners.

Thirteen names carry a strong distribution tag. The neocloud miner — up thirty and a half percent, the strongest name on the board — is one of them. So is the inspection-equipment maker, up fourteen and three-quarters. So is the infrastructure contractor, up seventeen and a quarter. Fifteen days of institutional supply, and the shares went up thirty percent anyway. That is the signature of covering, not of buying.

Four. Almost all of the darkpool “buying” is a single cross, and the memory buying came on a shrinking tape.

The institutional off-exchange file shows one hundred ninety-one billion dollars across three thousand one hundred thirty-three names, with a net of minus fifteen and a half billion. Split it by how many separate prints each name actually produced and the picture inverts:

Six billion dollars of apparent buying in the biggest memory name came from four prints. One and two-tenths billion in the storage name came from one print. One and one-tenth billion in the equipment maker: one print. Eight hundred forty-one million in the analog chip name: one print. Eight hundred sixteen million in the foundry: zero darkpool trades and three blocks. And the number the working brief did not carry, which is the decisive one: the memory buying printed on falling volume. The biggest memory name’s four billion dollars of “accumulation” arrived on a tape sixteen percent thinner than the day before. The storage name’s came on a tape down forty-three percent. Genuine accumulation expands the tape. This contracted it.

Five. Nobody bought upside. Not anywhere.

This is the finding that should end the argument, and it requires stripping the options file properly before it is visible. Of twenty-one and a half billion dollars of gross premium across forty-three thousand nine hundred forty-four prints, three and six-tenths billion is deep-in-the-money stock substitutes and three and two-tenths billion is matched call-and-put pairs with no direction. What is left — the actual directional residue — is fourteen and eight-tenths billion. Decompose that by cohort and by side:

Selling a put is not the same statement as buying a call. Selling a put says I do not think it goes much lower and collects money for that opinion, with the entire upside capped at the premium received. Buying a call says I think it goes higher and pays money for unlimited participation. On the biggest up day of the month, in the three cohorts that did all the moving, the professional side did the first and refused to do the second. That is a market closing shorts and harvesting inflated premium, not a market taking a position.

Six. Somebody bought five hundred and fifty-nine million dollars net of index downside — on the way up.

Three blocks, three separate expiration dates, forty thousand contracts each, all opening positions:

The September position sold and the October position bought are the same strikes rolled two weeks to the right — a deliberate put calendar for a sixty-two million dollar debit, and the roll harvested nine points of skew as well as two weeks of time. The resulting posture is precise and it is dated: protected into late August, deliberately open through the September quarter-end, protected again into mid-October. A four-hundred-lot version of the August position had been worked at 09:34:50, two hours and sixteen minutes before the real block — the desk tested the size of the market before it committed.

And one coincidence deserves to be stated flatly. Friday’s two-sigma downside for the S&P index is 7336.95. The lower strike of that August hedge is 7350. Whoever bought it is not hedging Friday. They are hedging a path.

---

The One Point Two Eight Billion That Is Not a Trade

At 14:40:45 — one second — eight legs printed in the S&P index, five thousand contracts each, at the two strikes that carry the largest standing open interest on the entire board. September calls sold at seven thousand, September puts bought at eight thousand, December calls sold at seven thousand, December puts bought at eight thousand, and the four smaller mirror legs. Total premium: one billion two hundred eighty-two million dollars. Five and nine-tenths percent of the entire day’s options tape, at a single timestamp.

This is a jelly roll: a synthetic-short calendar. It is a pure financing structure with essentially no equity exposure — a way to move a position’s funding from one expiration to another, executed at the strikes where the open interest already lives. The September seven-thousand put alone holds two hundred seventy-eight thousand contracts and one and seven-tenths billion dollars of notional; the December seven-thousand put holds one hundred fifty-one thousand contracts and two and three-tenths billion.

Any dashboard showing the S&P index with roughly five billion dollars of flow on Thursday is counting this, and it means nothing directional whatsoever. The index accounted for eight and one-tenth billion of the twenty-one and a half billion tape — thirty-seven percent — and one and three-tenths billion of that is a funding trade.

---

Microsoft — Fifteen Percent Up and the Most Distributed Name on the Tape

This is the highest-conviction single-name call in this report, and it is a sell. The software giant rose fifteen and a half percent on Thursday to four hundred fifty-one dollars and ten cents. Three independent measurements say that move was sold into.

Second: the institutional off-exchange tape. Minus four and twenty-seven hundredths billion dollars, on twenty separate prints — it clears the thick-tape gate — with volume up ninety-two percent day over day. That is the only print-count-validated, volume-expanding distribution in a major name anywhere in the file. Third: the fifteen-day ladder tags it distribution.

And the zone map is the most extended on the board. The close of four hundred fifty-one and ten is fourteen dollars and eighty-three cents above the zone ceiling of four hundred thirty-six twenty-seven. Upside room is negative three and three-tenths percent — there is no room left, the price is already through the top. Downside room is minus seventeen and six-tenths percent. The zone width is two hundred eighty-five, by a wide margin the most stretched reading in the document. Above the close sits a call wall of sixty-seven thousand contracts at the four-sixty strike for August expiration, ninety-five million dollars of standing supply nine points away.

Then it fell one and one-tenth percent after hours — the only major name red on a night when Amazon rose nine.

The counter-evidence, named rather than buried: thirty million dollars of near-dated Microsoft upside was genuinely bought the same session — seven thousand August four-thirty-five calls for seventeen million at 10:45, and nine thousand eight hundred ninety-five September four-sixty-five calls lifted at the ask for twelve and eight-tenths million at 12:06. The distribution call rests on the deep-in-the-money roll and the volume-validated tape, not on a claim that nobody bought calls.

---

Rotation Leaders Underneath the Surface

This is the question the report exists to answer, and the screen behind it now has three gates rather than two-and-a-half. First, a thick tape: at least fifteen separate institutional prints with a positive net, and day-over-day volume that expanded — the addition this cycle, forced by the fact that Wednesday’s four leaders all failed. Second, price participation. Third, structure-adjusted options flow that is positive after matched pairs and stock substitutes are stripped, with a strong preference for bought upside over sold downside.

Clears all three gates — two names in three thousand one hundred thirty-three

The Korea position is the single cleanest institutional footprint in the entire file, and the reason it exists is in the news layer rather than the tape. Assets in Korean leveraged exchange-traded funds have fallen thirty-seven billion dollars — about seventy percent — from an all-time high set barely thirty days ago, after a five-hundred-percent surge since the start of the year. That is the forced selling. What the tape shows is somebody on the other side of it.

The options in that name need naming honestly, because they look bullish and are not. At 09:43 somebody bought twenty thousand January-2028 forty-five-strike calls for forty-four and six-tenths million and forty thousand hundred-strike calls for forty-three and eight-tenths. Thirty-nine minutes later the same structure was sold one strike lower: twenty thousand forty-strike calls for forty-eight and three-tenths million and forty thousand ninety-strike calls for fifty million. They received ninety-eight and three-tenths million, paid eighty-eight and four-tenths, and extracted nine and nine-tenths million dollars in cash while moving a one-by-two ratio up five and ten dollars in strike. They also paid above-the-money volatility on the buys and accepted below-the-money volatility on the sales. That is a roll executed at a profit by somebody who already owned the position. It is not a purchase. The share accumulation is real; the option activity is portfolio maintenance. Both things are true and only the first one is a signal.

Clears two of three — the watch list

The data-centre analytics name, on three prints but fourteen blocks, plus six hundred twenty-four million net, an emerging accumulation tag, and the only megacap-adjacent name in the file combining an accumulation pattern with a positive net — it fell six-tenths of a percent on the day, which is exactly the profile of something being bought quietly rather than chased. The bitcoin-mining data-centre operator on fourteen prints and twenty-two blocks with volume up one hundred forty-two percent. The largest neocloud on fourteen prints and fourteen blocks with volume up thirty-eight — though its ladder says distribution, which is a genuine conflict. The coffee chain on five prints and six blocks with four hundred fifty-five million net and volume up one hundred ninety-three percent, on a name that has been left for dead. The California bank on three prints and thirteen blocks with four hundred eight million net. And the cooling-and-controls name held over from Wednesday, on four prints and eight blocks, plus sixty-seven million, up two and a half percent with its accumulation tag intact — the only survivor of Wednesday’s secondary list.

The names to be short or simply absent, ranked

---

Sector by Sector

Percentages of the tape below are of the fourteen-point-eight-billion-dollar structure-adjusted residue, not of gross premium.

Memory and storage — 10.9% of the tape, and the entire net is written premium

The complex rose eleven to twenty-six percent and not one name recovered its ten-session peak. The biggest memory manufacturer is still eleven and a half percent below its high of eight sessions ago and, at its overnight print, exactly back to where it closed six sessions ago. The storage name is twenty percent below its peak and, even after an eight and six-tenths percent after-hours print, still three percent below last Friday’s close. Only the drive maker carries an accumulation tag, and only the basket clears the tape gate. Both sides of the options book were net sellers.

Two enormous volatility purchases landed in the last thirty-five minutes, and they are the most expensive positions on the board:

The Korea memory proxy is where the downside is actually being paid for. Fifty-six thousand August tenth one-hundred-fourteen puts for eleven and two-tenths million at one hundred thirty-eight implied, against an at-the-money line near one hundred seventeen — a one-point-one-eight-times skew ratio, the only steep put skew anywhere in the complex. That name rose seventeen and a half percent on the day and its options net was minus fifteen and nine-tenths million. The fade grew into the strength.

Semiconductors — 11.5% of the tape, and zero net call buying

Calls net minus twenty-seven million. The entire ninety-seven-million-dollar bullish headline is put-selling. The best chip designer rose two and sixty-five hundredths percent, is tagged emerging distribution, and is pinned to a one-ninety-five strike carrying eighty-four and a half thousand contracts of standing September call open interest. Two structures define the professional stance: the equipment maker’s September three-forty put, seven thousand five hundred lots, sold at the bid for forty-five and three-tenths million and not tagged opening — a large bearish position being closed into the squeeze, not a bullish one opened. And the inspection-equipment name’s September two-twenty-four put, four thousand lots, sold at the bid for nineteen and three-tenths million at sixty-three implied against an at-the-money sixty-seven — downside written below the money. At the fund level: the chip fund’s January-2028 five-hundred put sold for twenty-six and eight-tenths million against an August five-fifty call also sold, and the other chip fund’s January-2027 four-ninety-five straddle sold outright for twenty-two and eight-tenths million. Eighteen months of semiconductor volatility, sold at the money, on the biggest semiconductor day of the month.

Neocloud, AI infrastructure and crypto — 4.0% of the tape, net short

Nineteen of twenty names carry bullish off-exchange flow and the ladder says two accumulating against five distributing. Calls were net sold by fifty-one and a half million on a cohort that rose seventeen to thirty percent. The largest print in the entire group: the AI-infrastructure name’s January-2028 four-hundred call, three thousand seven hundred eighty lots, sold at the bid for twenty-four and six-tenths million, at exactly the at-the-money line, on a plus-twenty-seven-percent day, in a name tagged distribution. The one contrary structure is the largest neocloud, where somebody bought December-2028 sixty and thirty-five puts for six and seven-tenths million — long-dated downside, purchased. In crypto there are zero accumulation tags across the brokerage, the two miners, the asset manager and the treasury company, on a day they rose seventeen to twenty-one percent.

Megacap — the dispersion was the story

The software giant plus fifteen and a half and distributing. The database company plus eight and a third with synthetic short built into it. The social network minus eight with an invalid zone. The search giant minus nine-tenths. The phone company minus one and four-tenths, then minus six after hours on a beat. The retailer is the only megacap carrying an accumulation tag — and even there the day’s own tape shows minus four and a half billion on thirteen prints with volume doubled, which is a genuine tension worth holding: accumulation on the fifteen-day ladder, distribution on Thursday’s tape, and a capital-spending raise the market paid nine percent for after the bell.

Software and IT services — the tag is killed

Seven names fell between four and six percent on a day the broad index rose one and seven-tenths: the workflow company minus four and nine-tenths, the design software company minus five and nine-tenths, the collaboration company minus five and nine-tenths, the human-resources software company minus five and nine-tenths, the consulting firm minus five and seven-tenths, the tax software company minus five and three-tenths, the Indian IT name minus five and two-tenths. Their zone maps carry thirteen to eighteen percent of downside room on widths of seventy-two to seventy-seven, which is extended in both directions at once. Wednesday’s promotion carried an explicit invalidation — a give-back of more than half the move in a single session. A seven-to-nine point relative give-back clears that in one print. The tag is dead.

And the structure that looked like counter-evidence does not survive reading its second leg:

Healthcare — downgraded one session after promotion

The sector fund fell one and six-tenths percent against a broad index up one and seven-tenths — a three-point relative loss the session after it was promoted. The largest pharmaceutical name fell four and a half and is tagged distribution. The consumer-health name fell three and seven-tenths and is tagged distribution. The tools company that led Wednesday’s list was flat with minus two hundred twenty-seven million on the tape. The whole sector printed eighty-one and a half million dollars of directional options residue — fifty-five hundredths of one percent of the tape — with a net of essentially zero. Three structures inside it, all of them maintenance rather than opinion: the biotech’s October three-ten call bought against its own August three-ten call sold, same size, same second, for a three-hundred-forty-thousand-dollar cost; the pharma name’s June-2027 sixty-two-and-a-half put bought against its December sixty-two-and-a-half sold, extending protection eighteen months for a million and a half; and the diabetes-drug maker’s August twelve-fifty-five put sold at the bid, which is a synthetic long. Six names in the cohort still carry accumulation tags, which is why this is a demotion to stabilization watch rather than a kill.

Energy — the third confirmation did not arrive, and the largest print was bearish

The entire sector printed sixty-nine and nine-tenths million dollars of directional residue — forty-seven hundredths of one percent of the tape — with a net of minus fourteen million, puts bought exceeding puts sold two to one. The single largest energy option of the day:

The sector fund itself printed twelve and four-tenths million and it was volatility, not direction: an October fifty-nine straddle bought, a September fifty-eight straddle bought, an August sixty call sold. The oilfield-services name printed four hundred thousand dollars of options and minus ninety-three million on the tape. Energy stays at Tier 2 candidate. It did not earn promotion and the option tape argues the other way.

Defense — one structure is the entire sector

Nineteen and one-tenth million dollars of directional residue across the whole complex — thirteen hundredths of one percent of the tape — and thirteen and two-tenths million of that is a single exchange-traded fund structure: a June-2027 two-thirty-five call bought against a two-fifty call sold and a two-thirty-five put sold, in two tranches at 13:11 and 14:59, all opening, a twenty-two-month call spread financed by a put sale for a net debit of roughly three hundred sixty thousand dollars. The primes are silent: the missile maker one million, the engine maker nine hundred thousand, the shipbuilder four hundred thousand, the tank maker zero. The communications prime fell eight and six-tenths percent.

Rates, credit and metals — three cross-asset tells

Somebody is short credit volatility out to 2027 and that is the most bullish single structure in the file. Against it, the long-bond fund shows a November eighty-six put sold and a November eighty-one put bought on nearly twenty-one thousand lots — a position for bonds lower and yields higher, in a zone whose width of seven and six-tenths marks the trend as dead.

And gold’s apparently bullish flow is net short. At 10:05:42, in one second, same ten-thousand size, identical ten-and-five-hundredths-million-dollar premium on each leg: a January-2027 three-fifty put bought at the ask against a January-2027 four-twenty call sold at the bid — a costless collar on roughly three hundred seventy-six million dollars of gold, floor three-fifty, cap four-twenty, and that three-fifty strike is a standing wall of forty-five thousand contracts. Then, from 14:23 to 15:39, a sustained purchase of deep-in-the-money December and January five-hundred puts — eleven thousand December contracts at around a hundred twenty-four dollars and three thousand January — on a name that closed at three hundred seventy-seven. Gold’s stock-substitute exposure for the day reached two hundred eighty-two million dollars and it is short. Zone upside room: eight-tenths of one percent.

---

Unusual Activity — Eight Structures Worth Reading

1. The delivery company’s roll — the same shape as the software giant’s, but bullish

Structurally identical to the software giant’s diagonal — and the opposite conclusion, because the new strike is further out of the money rather than deeper in. One desk is stepping down risk and taking money off; the other is stepping risk further out and paying at-the-money volatility to do it. The shape of a roll tells you nothing. The direction of the strike tells you everything.

2. The data-centre REIT’s single print

A November one-thousand put, two thousand five hundred lots, sold at the bid for thirteen and a half million dollars, opening, in one print. A bullish data-centre position expressed as written downside — in a name whose own fifteen-day ladder says emerging distribution. Worth watching precisely because the two disagree.

3. The Nasdaq fund’s 2027 call — the only large bought upside in the file

Eight months of Nasdaq upside for twenty-four volatility points, financed entirely by selling every near-dated wing. This is the whole bull case in one structure, and note what it is not: it is not a bet that the next month goes up. It is a bet that eight months from now is higher, funded by an explicit willingness to be short the next four weeks.

4. The phone company’s pre-print straddle

At 11:35:35, two hours before the largest consumer print of the quarter, thirteen thousand September three-thirty calls and thirteen thousand identical puts crossed for thirty-five and one-tenth million dollars, at 26.88 and 28.15 implied against an at-the-money line of 27.53 — magnitude bought at par, no wing premium paid on either side. The stock fell six percent overnight. That position was available to anyone reading the tape at lunchtime and it cost no volatility premium to own.

5. The database company’s synthetic short, built into a plus-eight-percent day

Twenty-nine and seven-tenths million dollars of August two-forty and two-fifty puts, every single leg tagged opening, printed between 14:23 and 15:23, at implied volatilities of one hundred thirty-five to one hundred sixty-nine against an at-the-money line of sixty-six and a half. Those ratios are not a volatility signal — they are the tell that these are stock substitutes rather than options. Roughly forty-two and a half million dollars of synthetic short exposure, opened on a day the shares rose eight and a third percent, in a name the ladder tags distribution with fourteen percent of downside room.

6. The neocloud miner’s options went the other way from its ladder

Nine and nine-tenths million dollars of August thirty-four and thirty-five puts sold — fifteen thousand at the bid, then eight thousand three hundred fifty at each of two strikes — at implied volatilities of one hundred twenty-one to one hundred twenty-seven against an at-the-money one hundred thirty-three. Downside written well below the money, on the strongest name of the day, which is simultaneously carrying one of only thirteen strong distribution tags in six hundred eighteen names. The option desk is constructive and the share ladder says fifteen days of supply. Structure before side: the ladder is a multi-day artifact and the option print is one day, so the ladder wins the tie on horizon. Do not chase it.

7. The defense call spread, twenty-two months out

Described in the sector section above, and worth repeating as a structure: the entire defense complex’s directional conviction for the day was three hundred sixty thousand dollars of net debit in an ETF, spread across two tranches. When a sector’s total expressed opinion costs less than a New York apartment, that sector is not being positioned in.

8. The tobacco put, and why it is the only real name on the cheapies panel

At 10:22:57, a September sixty-two-and-a-half put, fifteen thousand lots, bought at the ask in a block for one million three thousand dollars, opening — on a name that fell nine and three-tenths percent and whose zone width of minus forty and seven-tenths makes its map directionally invalid. Alongside it, a small-cap fund August two-seventy-seven put on thirty-seven thousand three hundred eleven contracts for six and three-tenths million. Those two are genuine. Everything else on that panel is not.

---

Do Not Trade These

The contract-idea and lowest-dollar-price panels rank by contract count, not by money at risk, which means sub-dollar lottery tickets own them every single day. A twenty-three-cent contract at a 195% out-of-the-money score is not a bargain — it is a contract whose implied volatility is being paid entirely for a tail the seller is happy to write:

And the same treatment for the off-exchange panels: six billion dollars of apparent buying in the largest memory name is four prints on falling volume. One and two-tenths billion in the storage name is one print on volume down forty-three percent. Eight hundred sixteen million in the foundry is zero darkpool trades. Eight hundred ten million in the mainframe company is zero. Seven hundred thirty-three million in the warehouse retailer is zero. On the red side, the same rule kills four and seven-tenths billion of “selling” in the phone company (six prints) and three and seven-tenths billion in the chip designer (three prints, on volume down sixteen percent).

---

Repairing a Broken Bearish Book

Near-dated implied volatility in this complex printed between eighty-five and one hundred sixty-three on Thursday, and that single fact changes what every repair costs. It also produces the most important observation in the volatility layer: across the entire memory and semiconductor complex, the out-of-the-money wings printed between 0.91 and 1.06 times at-the-money volatility. There is no skew premium being paid for downside in the names that just fell forty percent. The whole surface is elevated and flat. That is what a market looks like when it has stopped discriminating between up-risk and down-risk and is simply paying for movement — which is exactly why shorting the shares here is not tractable and why the professional money went short the volatility instead.

If you are long puts, deep underwater

Do not close them into this volatility. You are short an asset that just got marked up fifty to a hundred volatility points, and because the wings are flat there is no skew premium to hand back — the entire loss sits in the level, and levels mean-revert. The repair is to sell a nearer-dated put against the position and convert it into a calendar, harvesting the front month, which is the most inflated part of the surface. This is precisely what the professionals did on Thursday: the Korea memory proxy’s August-28 one-ten put bought against its August-7 one-twenty-five put sold, and the index desk’s September wings sold at 1.25× against October wings bought at 1.16×.

If you are short in-the-money call spreads

The worst-damaged structure in the file. The repair is not to roll up and out into the same volatility you are already short. Roll out only, keep the strike, take the calendar credit. The term structure in the memory names is inverted enough that an eleven-day option at one hundred sixty-three volatility comfortably funds a September option.

If you are short naked calls

Cover the gamma, keep the vega. Buy the nearest-dated call to flatten the tail that can actually hurt you, and re-sell a further-dated one at a lower absolute volatility. The institutional version of this printed at 10:07: an eighteen-month at-the-money straddle in the chip fund, sold outright for twenty-two and eight-tenths million, rather than fighting the front month.

If you are short stock

This is the position with no cheap repair, and it needs a rule rather than a fix. The borrow and the dealer gamma are both against you, so size is the only lever you have. Reduce to a size that survives another thirty percent in the weakest names, and express whatever view remains through the sector funds — where volatility is forty-eight to fifty-seven — rather than the single names, where it is one hundred nine to one hundred sixty-three.

Re-entry conditions, stated as conditions rather than prices

Timing a short re-entry by price in a complex printing one hundred sixty-three volatility is not a tractable problem, so the report does not pretend otherwise. Four conditions, in order of usefulness:

1. Volatility normalization with the wings still flat. Front-month memory implied back below sixty while the wings remain at 0.95 to 1.05 times at-the-money. Elevated-and-flat is a squeeze. Normal-and-flat is a range. Normal-and-steep is a top. Wait for steep.
2. Participation from the best chip designer. The short re-entry is not valid while it is the laggard. If it leads a rally, the squeeze has become a trend and the bearish thesis is simply wrong. If it keeps lagging a rising complex, the squeeze is intact and every new high is a better entry.
3. The behaviour of the August-28 block. That half-billion-dollar position is the single most informative object on the board. If it is rolled up or added to on further strength, the professional book is still short the path and the re-entry is live. If it is closed into strength, the hedge demand that defines this entire regime is gone and every bearish structure in this report needs re-deriving from scratch. Watch the 7400/7350 line specifically.
4. Volume expansion on a down day. Every distribution name on Thursday — the software giant, the used-car retailer, the two index funds — printed its negative net on expanding volume. The confirmation that distribution has become supply is a red session in which those same names print fifteen or more crosses with negative nets and volume up again.

---

Tier Board — What Changed

SOFTWARE AND IT SERVICES: Tier 3 emerging → KILLED, back to neutral-watch. The pre-registered invalidation fired in a single print — seven names down four to six percent on a green index, with thirteen to eighteen percent of downside room on extended widths. The ETF structure that looked like counter-evidence is a put calendar, protected only through August.

HEALTHCARE AND DEFENSIVE QUALITY: Tier 2 → DOWNGRADE to stabilization-watch. A three-point relative loss the session after promotion, distribution tags in the two largest constituents, and a sector options footprint of fifty-five hundredths of one percent of the tape with a net of zero. Six accumulation tags remain in the cohort, which is why this is a demotion rather than a kill.

ENERGY: Tier 2 candidate HELD — third confirmation NOT met. Forty-seven hundredths of one percent of the tape, a net of minus fourteen million, and the sector’s largest option of the day was an eighteen-and-a-half-million-dollar December put purchase in a refiner the ladder calls accumulation. The fund’s own flow was two straddles — volatility, not direction.

KOREA: CLOSED → RE-OPEN as Tier 2 LONG, first confirmation. The only name in three thousand one hundred thirty-three with a positive net on a thick, expanding tape, and an option structure that is hedged near and long into 2027. Confirm on a second session of thick-tape accumulation; invalidate on a close back below Wednesday’s low.

MEMORY AND CHIP EQUIPMENT: REDUCE HELD, sixth session — do-not-short caveat STRENGTHENED AGAIN. Nothing recovered its peak, the whole bullish net is written premium, and the wings are flat at 0.91 to 1.06 times. The constructive expression remains the basket, and the basket is now gate-confirmed.

NEOCLOUD AND THE AI-FINANCING SLEEVE: AVOID HELD, seventh session, WIDENED to include the squeeze winners. The cohort rose seventeen to thirty percent and printed fifty-one and a half million of net call selling. The instruction remains absence, not shorts — and Thursday is the proof of why that distinction matters.

CRYPTO: WEAK LEG HELD, no dip-buy. Zero accumulation tags across the entire complex on a twenty-percent up day.

THE BIGGEST CHIP NAME: REDUCE FURTHER → HOLD, reduction complete, do not add. Two and sixty-five hundredths percent is a failure to participate, not a recovery. Pinned to its strike with a hundred million dollars of standing call open interest, and tagged emerging distribution.

THE SOFTWARE GIANT: NEW TAG — REDUCE, immediate. Three independent confirmations in one session, which is this framework’s threshold for an immediate tag rather than an emerging one.

DEFENSE: NEW TAG — DORMANT. Insufficient participation to tier. Nineteen million dollars of expressed opinion across an entire sector, and the largest prime printed nothing at all.

---

The Hedge Calendar — Deferred Three Cycles, Built Now

The August-21, August-28, September-30 and October-16 structure is now explicit and dated, which makes the calendar constructible rather than speculative. Three dates matter more than the rest.

Diary these three: August 10–11 when the volatility cluster expires, August 28 when the protection line comes due, and September 30 — the gap somebody deliberately paid to create.

---

Bottom Line

Convergence reads net minus six. On the bear side: a hostile Fed with three dissents for a hike and no pause language, second-quarter output at one and a half percent against two expected, five hundred fifty-nine million net of index downside bought on a green day in a dated ladder, twenty-four and a half billion of print-count-validated off-exchange distribution against thirteen and a half billion of single-cross noise, an institutional share ladder that finished exactly even on a plus-three-point-three-percent Nasdaq, and the fact that the entire bullish options net in semiconductors and memory is written premium rather than bought upside. On the bull side: a capital-spending guide the market paid nine percent for, a supply-clearing block that removed the marginal seller permanently, a hundred thousand contracts of credit protection sold out to 2027, seventy-five million dollars of 2027 Nasdaq calls bought, and mechanical stretch. Four of those five bull inputs are mechanical or single-print. Only the capital-spending guide is a fundamental.

So the answer to the question this report was commissioned to answer is in two parts and they do not agree with each other. Wednesday was the low, in the specific sense that the forced seller is permanently gone and the position that was liquidating has already changed hands. The correction is not over, in the specific sense that nothing has re-rated: the software giant is fifteen percent higher and simultaneously the most distributed name on the tape, the database company is eight percent higher with forty-two million of synthetic short opened into it, the social network is sixteen percent below its peak with three hundred million of synthetic short on top of it, and the entire memory complex is still eleven to twenty percent below where it traded eight sessions ago while paying a hundred-plus volatility for the privilege of finding out what happens next.

The knife stopped falling. It has not turned into a floor. It turned into a range — and the professionals are trading that range from the short side of volatility and the long side of protection.

For a stock picker, that translates into an unusually short and unusually honest list. Two names in three thousand one hundred thirty-three clear a thick tape with expanding volume and a positive net: the Korea equity fund and the memory basket. Behind them, six names clear two of three gates and are worth watching rather than buying. Against them, the reduce list is the largest it has been in this cycle and it is led by the stock that just rose fifteen percent. The trade is not “buy the bounce” and it is not “short the squeeze.” It is: own the two names where institutions are actually accumulating on an expanding tape, be absent from the complex that rose thirty percent on nobody’s bid, reduce the megacap that ran through its ceiling, and let the August-28 hedge tell you when the regime changes.

---

Key Levels for Friday

The mechanical lines first. The zero-day gamma panels put the index fund’s positive shelf at 732–738 with the largest single bar at 738 and a wall of negative gamma from 740 to 744 — the fund closed at 741.69, inside the amplifying zone. The Nasdaq fund’s positive shelf runs 675–685 with its largest bar at 681, and negative gamma from 685 to 690. The S&P index has a single positive bar at 7420 against negative gamma stacked from 7430 to 7460. Dealers dampen below 738 on the index fund and 685 on the Nasdaq fund, and amplify above.

The zero-day put wall is the level that matters most. The Nasdaq fund’s 680 and 681 strikes carried four hundred eighty-seven thousand contracts of combined volume on Thursday, the largest concentration on the board, and the standing August 680 put open interest is another seventy thousand contracts. The fund closed at 683.55 — two and a half points above that wall. Its zone midpoint is 688.02, and the overnight tape cleared it.

Statistical guardrails from the forward expected-move table. S&P index 7437.63 with a daily move of 50.34: one-sigma 7487.97 and 7387.29, two-sigma 7538.31 and 7336.95. Index fund 741.69 ±5.23: 746.92 / 736.46, two-sigma 752.15 / 731.23. Nasdaq fund 683.55 ±9.01: 692.56 / 674.54, two-sigma 701.57 / 665.53. Small-cap fund 292.59 ±2.58.

And here is the number that should frame the entire session. The Nasdaq fund traded 691.62 overnight against a one-sigma upper of 692.56 — eighty-nine point six percent of Friday’s entire expected move was spent before the opening bell. The index fund has spent sixty-four percent. The biggest memory name has spent ninety-one. The retailer, the analog-adjacent chip name and the second chip designer have already printed through their one-sigma uppers. A band that is nine-tenths consumed pre-open is a band that has to be re-priced inside the session, and the modal outcome is a give-back rather than an extension.

Single-name levels worth having on the screen. The software giant: zone ceiling 436.27 against a 451.10 close, with 67,380 contracts of one-sided August 460 call open interest nine points above — no matching put line at that strike, so it is supply and not a financing structure — and the price is fifteen dollars above the top of its own map. The best chip designer: pinned at 195.04 against 84,580 contracts of September 195 call open interest — and that line is one-sided rather than a financing box, since the matching put open interest at the same strike is negligible, so it functions as genuine supply. Zone upside 9.9% and downside 3.6%. The retailer: zone midpoint 237.75, ceiling 254.94, and the overnight print at roughly 256.67 is through both. The phone company: zone floor 315.70 and a one-sigma lower of 320.63, with the overnight print at roughly 313.19 — through both, three dollars from its two-sigma floor. Gold: the collar’s floor at 350 and cap at 420, with only eight-tenths of a percent of zone upside room from 377.16.

Invalidations, spaced in volatility units. The bull case is dead on a Nasdaq-fund close below 674.54 and confirmed dead below 665.53. The bear case is dead on a Nasdaq-fund close above 692.56 and confirmed above 701.57. The credit re-arm is a high-yield close below 79.05 — the zone floor, thirteen cents from Thursday’s close of 79.47, and note that somebody just wrote a hundred thousand far-dated puts against that level. The rate channel takes over on a long-bond close below 82.44. And the single most important invalidation is not a price: it is the August-28 block. Rolled up or added to, the regime holds. Closed into strength, and this entire report needs re-deriving.

Calendar. Friday is month-end and brings the employment-cost index and the two remaining oil majors, with the chip-fund expiration in the background. Then August 3rd–7th for the space company’s print and lock-up, August 10–11 for the volatility cluster, August 21 for the largest standing options book of the cycle, August 28 for the protection line, and September 30 for the gap somebody paid sixty-two million dollars to create.

---

Sources: Live Options Flow 0730 (43,944 prints / $21.58B gross, 07:49:15–16:59:55 ET — full three-layer structure decomposition: deep-in-the-money stock substitutes 2,351 legs / $3,601M, matched call-and-put pairs 676 legs / $3,158M, directional residue 40,917 legs / $14.83B = 68.7% of gross; per-symbol side decomposition with calls and puts reported separately and never netted; size-weighted at-the-money volatility surface computed per symbol per expiry for every structure cited) · Darkpool Market Summary 0730 (3,133 names, $191.62B, net −$15.53B, with the print-count decomposition run natively off the trade-count and block-count columns and a day-over-day volume overlay added this cycle) · options dashboard 0730, twelve panels read as images: zero-day flow and gamma for all three index products, market delta exposure, flow map by expiration, flow timeline, dealers diary, call and put chains across four pages, and the contract-ideas panels · Options Market Summary extracts 0730: unusual contracts, cheapest options with out-of-the-money scoring, and large out-of-the-money open interest · ZONE DOCUMENT 0731 and Zone Visual 0731 (142 rows, built off the 07/30 close) plus the daily expected-move table for 07/31 with one- and two-sigma bands for fourteen index and futures products and sixteen single names · recon 2026-07-30 nightly pipeline: 618-name census with per-name price, off-exchange flow classification, fifteen-day whale-ladder pattern, gamma exposure, dealer positioning and fifteen-day flow history — every percentage in this report is read off that census and nowhere else · prior cycle: the 07/29 analysis file, the 07/29 published report, rolling tracker v59, regime snapshot 0729 · user-supplied Thursday event layer (the block sale and its counterparties, both after-hours mega-cap prints with full line detail, the electric-vehicle maker’s China separation report, the day’s ten most active options, model pricing cuts, the second-quarter output and inflation prints, Korean leveraged-ETF asset destruction, two Chinese gigawatt data-centre projects, the search giant’s backing of a Texas AI campus, the European Union gigafactory tender, and the retailer’s capital-spending raise). Declared inventory gap: the Market Commentary corpus holds 361 files whose newest are stamped 0728–0729; no 07/30-dated commentary is present on disk, so the narrative layer in this cycle rests on the user-supplied event layer and on the tape. Nothing in the transcription workspace was read or touched, and no file anywhere was moved, renamed or reorganized. Analysis file: the 07/30 verification layer — inventory gate satisfied, citation gate eight of eight with zero warnings, structure gate satisfied. Two figures carried in the working brief were corrected against the tape rather than repeated: the 14:40:45 index structure is eight legs and $1.282B (not seven and $1.53B), and the median single-session retracement is 54% (not ~31%).