Daily Report — 10/22/26 · “The AI Repair Failed and the Money Rotated, Not Left — What Friday and Next Week Need to Show”

Thursday in one line — the AI repair failed on spending news, and the equal weight rose while the Nasdaq fell

The S&P 500 (SPX) closed at 7,663.43, down 1.21%, under the 7,713.88 daily floor that our 10/21 edition named as the line that undoes Wednesday’s repair — so the repair is undone. The selling was set up before the open: META faded its after-hours gain on its spending outlook and its plan for a non-voting share class to fund it; TSLA and IBM traded lower after their reports. Jobless claims printed 194K at 7:30 Central, the lowest of the 13 readings on our record, so this was not a labor scare. By the 9:15 cut our AI basket was down 6.29% equal-weight with 0 of 33 names higher. The index reached Silva’s net gamma put wall at 7,650 by 10:15, held it through 12:45, then broke it and the zone floor to 7,637.36 at 14:30 as META’s spending news carried the selling into power equipment (GEV about -11.7%, VRT, ETN). It recovered 26 points into the close. VIX touched 20.31, over its zone ceiling 19.68, and closed back under it at 19.03.

This was a split day, not a liquidation. Of the 651 stocks and funds we track, 292 rose and 358 fell; the median one lost 0.74%. 148 fell 5% or more — the second-most of the 53 sessions on our name ledger — and they were the AI hardware, power-equipment and speculative names. On the other side, RSP rose 0.73% while QQQ fell 2.86%, a 3.59-point gap and the widest equal-weight lead in those 53 sessions. Staples (XLP) rose 3.31%, their best session on the ledger, while SMH fell 5.99%, its worst. Regional banks (KRE) rose 2.53% and the Dow closed flat after printing over its October ceiling. When banks lead and claims are low, the money is rotating inside the index, not running from it.

The AI repair failed on all three names. SMH closed at 523.35, 32.84 under its October floor 556.19; MU at 856.49, under 911.47; AMD at 507.93, under 536.09. INTC (85.49) and DRAM stayed broken. Every one of our 19 bearish group calls regained its price input on the close, and two retired on evidence gone (Robotics & Automation, Transport & Logistics). SMH now sits 9.30 over its 200-day average (514.05) and 19.13 over its July low (504.22) — the two lines we registered on Thursday to confirm the group’s lower-high, lower-low trend on the fund itself.

The regime under it — the biggest one-day bond rally in a year, credit calm, and the dollar block clear

The Fed is still tightening, and Thursday’s bond rally did not rescue the AI multiples. The Fed raised its rate a quarter point to 3.75% to 4.00% on 9/16; the next decision is Wednesday 10/28 at 13:00 Central, and the economic calendar’s consensus is a hold. The 10-year closed at 4.961%, down 15.8 basis points — the largest one-day fall in the 258 sessions of our daily record and its first close under 5% since 9/23 — with the two-year down 16.0 to 4.648% and the 30-year down 12.4 to 5.286%. Most of the fall was in inflation compensation: the 10-year breakeven slipped from 2.37% to 2.32%, and the 5-year inflation-protected auction cleared at a real yield of 2.367%, up from 1.955% last time. The real 10-year yield, 2.76% for 10/21, still sits well above the breakeven. When the real yield runs above the breakeven, the bond market is doing the Fed’s tightening for it, and that caps what investors pay for high-multiple stocks — which is why lower nominal yields did not help them on Thursday. The long-bond fund TLT rose 1.87% to 81.74, over its weekly ceiling 81.36 and its October ceiling 81.07: a continuation, and its input is the rate move itself as money moved into duration.

Credit did not transmit the equity selling. The high-yield spread printed 3.27 for 10/21, still over the 3.12 line we watch but 6 basis points tighter. HYG rose 0.16% to 77.10, at the middle of its Friday zone 76.75 / 77.07 / 77.38, and 0.71 over the 76.39 line we registered Thursday as a liquidity-drain marker. KRE closed at 72.41, over its weekly ceiling 72.16. We registered five lines on Thursday that would mark a liquidity drain rather than a rotation — HYG under 76.39, the yen through 155.27, the three-month implied correlation index over 20, a weekly fall in KRE with HYG, and gold and bonds falling with stocks. None fired: the correlation index closed at 14.56, up 13.5% on the day but well under 20, and bonds rose while stocks fell.

The dollar block on metals is clear, and gold fell for a reason the regime does not explain. The dollar index closed at 101.62, about 23% of the way up its Friday zone 101.359 / 101.92 / 102.47, with a range reading of 21. Our strong-dollar block on new bullish metals structure needs the dollar over 100 with a range reading over 40, so it is off, and the dollar is not extended in its zone. GLD fell 2.63% to 369.00, under its weekly floor 374.00 and its Friday zone floor 370.14, with the dollar flat and yields down. December crude rose 2.4% to 99.47 after a 100.17 high. Gold falling with bonds rising is the one early liquidity marker on the board; until credit confirms it, it is a fall in a crowded holding, and our Gold & Silver Miners bearish group call governs GDX (81.69, under its weekly floor 83.69).

Sentiment barely moved. The FOM sentiment gauge printed 48.8 (neutral) for 10/22, down 4.7 on the day and 0.4 over five sessions — a small move on the AI basket’s worst day because the gauge reads the average stock, and the average stock held up. It is 33.8 points over the capitulation floor at 15 and adds nothing to our count; no bottom signal is in it.

The four timeframes — the daily floors broke, every swing band held, and the week ends Friday with the index on its anchor

The quarter and the month are untouched. The S&P sits almost exactly on the anchors of its quarterly band for the fourth quarter (7,089.48 to 8,213.60) and its October band (7,363.93 to 7,939.15) — Thursday’s fall erased the month’s gain, nothing more. A selloff that leaves the swing bands untouched is a move inside the regime, which is why we do not call an index direction from it. The Nasdaq fund is closer to its edge: QQQ traded 701.90, through its October floor 702.09, and closed 2.38 over it; the Nasdaq 100 came within a tenth of an expected move of its own October floor (28,843.31). A touch of a monthly floor that recovers carries no measured basis for a bounce, because we have no tested rate on that cell; a QQQ close under 702.09 would be the first monthly-band break of this leg.

The week used up its gain and has one session left. The weekly band runs 7,490.45 to 7,828.43, anchored on last Friday’s 7,659.44 close, and the S&P closed four points over that anchor: four sessions of round trip, 173 points over the floor. QQQ sits 17 points over its weekly floor 687.40. The small caps went the other way: IWM traded through its weekly ceiling 277.13 for a third session (high 277.28) and closed back inside at 276.67, which leaves the weak fade lean we have carried all week, graded on a Friday close under 276.67; a close over 277.13 is a weekly break instead. That lean missed twice already this week, so we read it as nearly no basis.

The daily band broke from the open. Thursday’s board ran 7,713.88 to 7,800.46; the index never traded inside it (its high was 7,702.94) and closed under the 2-sigma floor too (7,670.59), after reaching Silva’s zone floor 7,642.39 and his net gamma put wall 7,650. In the current tightening regime, an index close under its daily floor has come back inside the next session 6 times of 8 since 9/16, a small sample measured on the indices only; that is the rebound setup, graded on a Friday S&P close over 7,663.43. It is not a buy signal: our reversion-buy gate stays closed while VIX sits over the middle of its zone and the index under the 7,712 gamma flip. Friday’s board is wider — 7,604.74 to 7,722.12, 58.69 points each way against 43.29 on Thursday — and Friday’s zone is 7,629.03 / 7,728.09 / 7,827.15 with a range reading of 10, a dead trend reading. QQQ’s Friday board is 693.98 to 714.96, zone 692.20 / 723.59 / 754.97; IWM’s 274.01 to 279.33.

What it does to the forward view. Because the swing bands held, a Friday bounce has room: the week’s ceiling is 165 points away and the month’s ceiling 276. Because the daily floor broke and the index closed under the flip, a failed bounce has room too: Friday’s zone floor 7,629.03 is 34 points down and the week’s floor 173. The bands price the size of the next move; they do not choose its sign. Next week’s band is not posted by either source yet: Silva’s weekly board comes after Friday’s close, and our option-implied weekly band re-anchors on Friday’s settle. An indicative width from month-end option prices puts next week at about 2.2% each way for SPY and 4.2% for QQQ, wider for the Nasdaq because the Fed and the megacap reports sit inside it.

Outside the index, the bands split the market in two. On the downside, SMH (523.35), DRAM (45.68) and the technology fund XLK (182.34, a first close under its October floor 183.73) closed under their October floors: a continuation of the AI groups’ trend, its input the day’s company news on AI spending, which fails only on closes back over 556.19, 51.89 and 183.73. AMD, MU and INTC sit under their October floors too, and META (648.85) and GOOGL (324.22) under their weekly floors; a single stock’s band break carries no measured basis for a bounce, so the group calls govern. Gold, silver and the miners broke their weekly floors with no regime input behind them. On the upside, staples (XLP 89.99) closed 3.13 over their October 2-sigma ceiling, health care (XLV 181.30) over its weekly 2-sigma ceiling, and banks, software, biotech and the Dow over their weekly or October ceilings — no measured basis for a fade on any of them. The long bond and the rate-sensitive funds (TLT, the utilities, real estate, homebuilders) are the one upside continuation with an input behind it: the rate move itself, which holds while TLT stays over 81.36 and fails on a 10-year back over 5.119%. Across our census, 217 names closed under their option-implied daily ranges and 183 over them.

The mega-caps and the roster — 186 names; AAPL was the only core name at a high, and the AI chain gave back Wednesday’s repair

The largest names that are not spending on AI held up; the ones that are, or sell to them, did not. AAPL closed at 348.22 (+1.23%), the only one of our sixteen core names at a 20-session closing high, inside its weekly band 333.36 to 353.56 and its October band 311.95 to 354.09, at 83% of its zone 335.31 / 343.07 / 350.82: inside technology, the rotation went to the company with the least AI spending to defend. NFLX rose 3.09% to 67.05, the best core move, at 23% of its zone 65.23 / 69.22 / 73.20 and inside its weekly band 63.92 to 77.16, which leaves it a trend name recovering from its report, not a bounce setup. MSFT fell 2.23% to 521.08, 3.29 over the floor of its zone 517.79 / 532.18 / 546.56 and inside its weekly band 512.63 to 545.47; it reports after the Fed on Wednesday 10/28, and its spending plans are the same question META raised. AMZN fell 4.57% to 250.38, 2.00 over its zone floor (zone 248.38 / 256.88 / 265.38), giving back Wednesday’s third close over its weekly ceiling 259.27; it reports Thursday 10/29. TSLA fell 3.08% to 374.58 after buying back its whole post-report gap in the first hour (open 374.03, high 388.86, through its weekly ceiling 382.86), at the middle of its zone 356.40 / 372.57 / 388.73.

META broke its weekly floor on its own report, and GOOGL followed it. META fell 6.98% to 648.85, 7.68 under its weekly floor 656.53 and under its zone floor (zone 655.22 / 706.58 / 757.93, a negative range reading), after a 639.69 low through its October floor 645.04; it closed 3.81 back over that line. The report itself was strong — revenue up 26%, a fourth-quarter guide over consensus — and the market sold the spending: a 2026 capital budget narrowed to $135B to $145B, the first negative free cash flow since 2012, and a non-voting share class to fund it. Its front-month options lost 29% of their implied vol on the day as the event premium came out. That is a continuation on its own report, and a close under 645.04 extends it. GOOGL fell 3.15% to 324.22, 4.08 under Silva’s weekly floor 328.30 and under its zone floor (zone 326.89 / 342.03 / 357.17, a negative range reading); GOOG fell 3.08% to 320.76, under its option-implied weekly floor 322.11 and its zone floor (zone 323.19 / 338.21 / 353.22). The two classes moved the same: one company, no class divergence, and both report Wednesday 10/28 after the close.

The chip names gave back the repair on their own October floors. AMD fell 8.10% to 507.93, 28.16 under its October floor 536.09 and 14.46 over its weekly floor 493.47, at 21% of its zone 475.79 / 551.28 / 626.77. MU fell 6.26% to 856.49, 54.98 under its October floor 911.47 and 37.80 over its weekly floor 818.69, at 27% of its zone 785.03 / 917.57 / 1,050.10; it traded 880.28 overnight as Korea’s memory names reversed. NVDA fell 4.40% to 212.21, 2.08 over its October floor 210.13 after a 211.26 low, at 26% of its zone 203.89 / 219.78 / 235.66. AVGO fell 4.31% to 323.24, 2.52 over its October floor 320.72, at 19% of its zone 312.29 / 341.28 / 370.27. TSM fell 4.52% to 412.35, 0.03 over its option-implied month-end floor 412.32, at 33% of its zone 386.78 / 425.62 / 464.45. INTC fell 8.04% to 85.49, still 13.09 under its October floor 98.58 — the deepest break of the sixteen, under it since 10/14 — and 2.50 over its weekly floor 82.99, at 29% of its zone 77.47 / 91.34 / 105.20; it reports Thursday 10/29. Every one of these sits inside our bearish AI basket group call, which retires only on an SMH close over 611.01. A single stock’s band break carries no measured basis for a bounce: the group trend is the read.

SPCX and MSTR, the event names. SPCX fell 9.14% to 144.75 on the day of its largest post-listing share unlock, inside its weekly band 131.99 to 160.07 and 9.95 over its October floor 134.80; it traded 148.58 overnight on reports that SpaceX is seeking 80,000 cell-tower sites and finalizing a $40B debt package priced at 7.75% to 8.3%. SPCX ZONE GAP: it has no zone row. MSTR fell 5.00% to 145.80, back under its weekly ceiling 149.50 after a tag of it, inside its October band 128.27 to 177.91, as bitcoin fell through its daily 2-sigma floor to about 81,450; the crypto group stays in a bearish call. MSTR ZONE GAP: no zone row.

The rest of the roster, grouped. The day’s reporters split by what they sell. AI suppliers that beat were sold: LRCX -11.76% to 230.15 (zone 220.10 / 258.66 / 297.21), NOK -14.81% on a data-center beat, CRS -20.62%, and after the close WDC traded 345.52 (-2.45%) on record profit and a higher outlook. Companies that sell to consumers, governments and investors were paid: LMT +6.96% to 589.91, at the top of its zone 515.25 / 552.82 / 590.39; PG +5.58% to 165.59 (zone 151.76 / 159.48 / 167.19); BX +8.38%; SPOT +11.10%. IBM, sold after hours on Wednesday, closed up 0.26% at 252.11 (zone 221.31 / 238.69 / 256.06). Power equipment took META’s spending news hardest: GEV -10.64% to 827.42, at the floor of its zone 821.42 / 919.92 / 1,018.42, with VRT and ETN behind it. Services and software caught the rotation: ACN +7.69% to 258.74 (zone 211.90 / 237.64 / 263.37), ADBE +6.07%, with our hourly cyber-and-software watch moving to a top watch as CHKP drew an exhaustion tag on a +4.7% day. The memory-and-semis bottom watch printed no bottom.

The sector map — the rotation of the rotation resumed at its widest margin on our record, with no bottom test and no credit stress

Thursday put back the order of 10/6 to 10/20 that Wednesday had briefly reversed. Rebuilt stock by stock from the 651 names we track, staples led (median +3.76%), then real estate (+3.16%), health care (+1.16%) and communication services (+0.80%); technology (-4.82%) and materials (-5.79%) were last. Nine of the 11 sector funds beat SPY, against three on Wednesday. Every one of the 40 chip names fell, 33 of them to 20-session closing lows. Thursday’s stock-by-stock moves ranked +0.70 with last week’s and -0.48 with Wednesday’s: the one-day repair was the exception, and the trend came back. RSP beat QQQ by 3.59 points (the widest of 53 sessions on our record), and staples beat SMH by 9.30 (the widest of 57).

Two sector funds disagree with their own stocks, and the stocks are the read. The utilities fund XLU rose 2.07% on its 17 regulated names, every one at a 20-session closing high, while the utilities group as a whole fell because GEV (-10.64%) and the merchant power names (VST, TLN, CEG) sold off on META’s spending outlook: utilities split, they did not fall. The materials fund XLB rose while the materials stocks we track are mostly miners, and those fell (3 of 32 up). The financials fund XLF rose on banks, payments and asset managers, while the bitcoin miners filed under financials fell 10% or more.

Each side owned a different half of the market. Of the 255 names down 2% or more, 132 sit in the AI complex and 45 in the speculative end; of the 168 names up 2% or more, 151 sit outside both — staples, health care, travel and consumer brands, regulated utilities, housing, data and exchanges, defence, banks. The bottom-up dark-pool sums say the same in block-trade dollars (closing prints, a description and never a signal): staples dark net +$4.00B and financials dark net +$4.36B against technology dark net -$3.41B.

This was a trend continuing, not a bottom test and not a de-grossing. The most-beaten third of the market closed 30 up and 186 down, and the third nearest its highs 162 up and 55 down. A crowded-longs unwind would have the strongest stocks falling hardest; a bottom would have the beaten names leading an up day. Neither printed. The selling went where the damage already was, inside our 19 bearish group calls, every one of which lagged the market the way it was called, and every fund’s retire line moved further away (XLI’s is the nearest, 1.76% over Thursday’s close).

How far it has gone, and how long these last. Since 10/6, SMH is down 17.26% while staples are up 10.01%, a 27.3-point gap in 13 sessions, with the two moving almost exactly opposite each other day to day. The defensive side is now stretched: XLP closed three October expected moves over its anchor and over its October 2-sigma ceiling (86.86), 0.02 under its highest weekly close of the year; XLV sits over its weekly 2-sigma ceiling (180.99). A stretched band on a sector fund carries no measured basis for a fade. On weekly bars, SMH’s down leg is 18 weeks old; the two prior AI legs ran 39 weeks (SMH, 2024-25) and 22 weeks (QQQ, 2025-26), and in both the defensive bid came only in the down legs. QQQ’s own leg is two weeks old. Two cycles is an anecdote, not a rate.

What would end it, in the order it would show. A big up day in the beaten groups led by quality — NVDA, TSM, AVGO ahead of the low-priced names — would be a real bottom test; Wednesday’s up day was led by the squeeze names and lasted one session. Then the group calls losing their evidence or their lines (SMH over 599.32 and 611.01, DRAM over 57.44, XLI over 169.67). On the other side, staples under 83.50 would end their leading state. Next week’s catalysts sit on both sides of the pair: GEV and VRT report before the Fed on Wednesday, MSFT, GOOGL and NOW after it, and a rise in yields would take away the one input that paid the defensives on Thursday — the 10-year traded 4.982% overnight.

Energy did not follow crude. December crude closed at 99.47, in the middle of Silva’s weekly band (93.78 to 104.30), and traded over 100 overnight. A breakout is a weekly close over 104.30; until then 100 is the middle of the range, not a line. Energy stocks fell with crude up: three of 12 producers rose, oil services fell 3.03%, and XLE lost 0.87%.

The dealer map — the first dashboard in four sessions: Friday’s biggest row is stock substitutes rolling off, and the index insurance moved past the Fed

The options dashboard is back, and its loudest panels are mostly plumbing. Tradytics’ Dealers Diary, which plots dealer delta by expiry, shows its biggest forward row on Friday 10/23: 6.46 billion of put delta. Our decomposition of the flow file matches that row almost exactly (6.47 billion of share-equivalent delta) to one afternoon’s deep in-the-money Friday puts — about $3.1B of premium between 1:42 and 3:09 Central in GLD, LRCX, ORCL, SMH, MU, AMD, NFLX, META and AVGO, priced almost entirely at intrinsic value. Those are stock substitutes, not bets: they carry delta, not conviction, and they expire Friday, so nothing in that row is a level for next week. Behind Friday the dealer book is light through the Fed (Monday 2.22 billion, Tuesday 1.89, the Fed day 0.34, Thursday 0.21) and heavier, mostly puts, at month-end (2.17) and in November (2.14).

The real protection was bought for Friday and month-end, and the index hedges were pushed past the Fed. Market Net Flow gave up calls and paid for puts all session. The Flow Timeline’s Friday line fell to a new low after a one-day rise on Wednesday, and so did its October 30 line; because the flow file for Friday shows real put buying, this is new protection, not old hedges netting out. In the index book, QQQ hedges were rolled out of Friday and month-end into November 6 and November 20 puts — past the Fed, the cloud reports and month-end — and SPY bought a Monday 755 / 750 put spread of about 80,000 lots after the close, across the Sunday Brazil runoff. The zero-day gamma book was negative at the close (the SPX panel’s largest strikes were 7660 and 7640, both negative), which is the side where dealer hedging adds to moves; it expired with Thursday and is history.

What it does to Friday and next week. Friday’s positioning is mostly expiring stock substitutes plus a one-day put bid, so the open, the flash PMIs and the overnight reporters should set the day more than any dealer level. Next week the dealer book is light into the Fed and the index insurance is dated after it, so positioning does not hold the index up or down into Wednesday 10/28. The largest S&P open-interest lines (December 7000 and 8000) are box-spread financing at parity — a loan, not a floor — and the one genuinely one-sided wall on the board is EWZ’s November 43 calls into the runoff.

The options tape — a 71% bigger file that was two-thirds plumbing, and a put lean built by a few tickets

The file nearly doubled, and most of the growth was financing and expiring stock substitutes. Total premium rose 70.9% to $24.54B, and 66% of it was structure. About 15,200 lots of the S&P’s December box spread rolled into September and December 2027 boxes at 5.02% and 5.08% simple — the market’s own borrowing rate, down from 5.23% for December 2027 on Wednesday, a financing trade and not a direction. The deep in-the-money Friday puts added another $3.77B of stock-substitute premium after 1:30 Central.

After structure, the equity tape leaned to puts, and a third of the lean is one ticket. Stripped of packages and stock substitutes, the single-stock tape’s directional residue was -$247.7M on $5.10B of sided premium — calls written over bought by $112.3M and puts bought over written by $135.5M, the largest such lean since 10/16 — built between 12:45 and 2:30 Central. Its single biggest line is PLTR: one 25,000-lot December 175 call block sold to the bid at 2:15 Central for $76.75M, 20.9 points in the money. That is a holder selling in-the-money calls or writing calls against stock; heavy call selling on a name that has run is a holder locking gains, never a short. Index options were 80% structure, and on the down day index puts were written over bought by $81.8M while the hedges that did trade were rolled down and out past the Fed. The defensive side barely touched options: TLT, XLP and XLV together printed under $80M, and the protection that did print was small-dollar credit and bank puts (HYG puts at 4.7 times their average volume, KRE November 65 puts to open). Our monetization test reads neutral across the file: holders writing calls against stock, not the harvest of a run.

Volatility and mood — the front month carries the Fed and the cloud reports, index calls price under the money’s vol, and correlation is climbing

The index curve is inverted at the front. SPX at-the-money vol for October 30 is 16.4, up 33% on the day, over 15.0 for January; QQQ’s is 29.2 over 22.7, the steepest index inversion on the board. The October 30 expiry holds the Fed on Wednesday and the cloud reports on Wednesday and Thursday, so the market is paying up for next week specifically. The November strikes repriced higher while the index fell (the put wing up 1.9 vol points over five sessions): that is genuine hedging demand, not the mechanical lift a falling index gets from its own skew. Index put wings price 1.30 to 1.46 times the at-the-money vol; call wings price 0.82 to 0.92 times. Measured by its own vol, index upside is the cheap contract and protection the dear one.

Correlation is rising as the AI trade unwinds. The three-month implied correlation index closed at 14.56, up 13.5% on the day and about double its June low. When correlation rises from a low, the dispersion position — SPX index vol sold against single-stock vol bought — is being taken off — consistent with the pair of AI longs against defensive shorts coming apart. It is a description of positioning, never a direction. Of the 49 names we can measure, 36 carry a front month at or above 1.10 times their back month, ten of them with no dated event we can name.

Unusual trades — the structures worth naming

SPX — the December box spread rolled into 2027 at about 5%, and index puts written into the fall

By structure this is one roll: about 15,200 lots of the December box spread were closed and reopened in September and December 2027 boxes within three seconds, at 5.02% and 5.08% simple. A box spread is a loan; its rate is the market’s own borrowing cost, now about 43 basis points over the two-year Treasury and lower than Wednesday’s 5.23% for the same date. Around it, index puts were written over bought on a down day. SPX closed at 7,663.43, at 17% of its Friday zone 7,629.03 / 7,728.09 / 7,827.15, so none of this flow sets a level; it is financing and premium selling.

PLTR — one 25,000-lot in-the-money call block to the bid, half the day’s line

At 2:15 Central one print sold 25,000 December 175 calls to the bid at 30.70, $76.75M, 48% of the day’s volume in that line. Structure before side: the calls were 20.9 points in the money at 60.0 implied vol (1.04 times the money), so this is a holder selling in-the-money calls or an overwrite against stock — never a short. Friday’s open-interest count settles which. PLTR closed at 196.19 (-2.74%), at 43% of its zone 183.45 / 198.24 / 213.03, just under the 197.01 line on the operator’s PLTR spreads.

MU and the AI complex — calls bought as a two-strike package on the worst day, and protection rolled down

By structure, MU’s January 1030 and 1070 calls traded as a same-side two-strike package of 3,611 lots a strike to open, both fills at the ask, about $27.1M — a package, so its premium is the sum, and long-dated upside was bought on the AI basket’s worst day. MU closed at 856.49, at 27% of its zone 785.03 / 917.57 / 1,050.10. In the same complex SMH’s October 30 530 puts were rolled into November 6 500 puts (9,990 lots), AMD’s November 550 puts traded in the money, and SOXL January 2029 puts were bought to open. A package bought and a hedge rolled lower are positioning on both sides; the bearish AI basket group call, not the options, carries the read.

QQQ — the year-end short call rolled 15 points closer, and month-end puts rolled into November

By structure, QQQ’s December 31 740 calls were bought back and 725 calls sold in one roll of 38,000 lots a leg: an overwrite re-struck closer to the price, a holder capping upside into year-end, not a bet on direction. October 30 700 puts were rolled into November 6 693 puts (20,000 lots) and October 30 690 puts were bought (28,716). QQQ closed at 704.47, at 20% of its Friday zone 692.20 / 723.59 / 754.97; the largest October 30 open interest sits in the 700 and 690 puts, both inside Friday’s daily band.

EWZ — a 110,000-lot call spread into Sunday’s Brazil runoff

By structure this is a December 46 / 50 call spread of about 110,609 lots, the largest single-name upside ticket of the day, bought as one package into the runoff; separately the November 43 call line, 809,199 contracts against almost no puts, is the one genuinely one-sided open-interest wall on the board. EWZ closed at 41.85, at 23% of its zone 40.75 / 43.18 / 45.60. The runoff sits inside the October 30 expiry, which is why its front vol runs about twice December’s.

The commentary — the desks agreed on what happened and split on what comes next, Silva’s lines all failed, and the Maverick read the rotation best

Mike Silva’s levels framed Thursday, and his forward lines failed. No Silva deck or video dated 10/22 was on the shelf, so we graded his 10/21 pair and his Thursday morning brief. His same-day gamma anchors framed the session: the first-hour high stopped just over his 7,700 same-day put wall and never regained his 7,712 flip, and the low tested his 7,650 net gamma put wall before the close came back over it. Every forward line he set for Thursday failed or went untested: SPY’s 768 “hold this low” broke from the open, QQQ’s series of higher lows broke, SMH never neared his 560. His regime label is built on two inputs — the S&P against his gamma flip and the Nasdaq Composite’s 10-day average — and on Thursday’s close both point back to his bearish-strong cell, one session after he moved to bullish-weak; that is our arithmetic on his definition, and his label is his to state in Friday’s deck. The one call still open is his weekly tag: SPY needs 2.3% down or 2.1% up on Friday to touch either edge of its weekly band (746.72 / 779.78).

The Maverick read the rotation best of anyone this month; his trigger was not the cause. He is a standing bear, and we use his framework, never his mood. His Thursday morning condition — “if oil trades above 100, the Nasdaq gives up 30k” — saw the Nasdaq leg happen while crude closed under 100, on AI-spending news rather than oil. His evening video says the rotation out of tech is back, any chip bounce is a fake until his oil-dollar-bond map leaves stagflation, and the shorts are crowded enough that he expects a bottom “soon” after one more flush that could run into the 11/3 midterms — the flush where he would sell index puts and buy dips. His downside levels are SPY 760, 750, 740 and the 200-day near 726.65, and QQQ’s 200-day near 676.89. On a Democratic sweep he expects the AI build-out names, the cloud spenders and the Musk companies to lose, and software, homebuilders, hospitals and dollar stores to win. We score his calls at their dates; none of it is a trigger for us.

The other desks split on next week, not on Thursday. Nearly every desk on the shelf read Thursday as the census did: money out of AI and power, into bonds, staples, health care, utilities and the equal weight. The bounce camp — Geeks of Finance (7,650 holds, a squeeze over 7,700), Krown (a bounce near 7,600), Mike Jones (“a tech wreck, not a market crash”), Darius Dale (a rebound into early 2027) — sits beside the continuation camp: SpotGamma’s Brent Kochuba (“if we get 7,600, then I think we get 7,450”), Ilya Spivak (the Nasdaq 100 toward 28,700 to 28,850, then lower), and Tim Knight, 140% short, whose one number on SMH (536) broke on Thursday. Their levels line up with ours: SPX 7,600 sits on Friday’s daily floor 7,604.74, Kochuba’s 7,450 sits under the week’s floor 7,490.45 and over October’s 7,363.93, and Spivak’s Nasdaq 100 range sits on its October floor 28,843.31. James at InvestAnswers is buying the dip in NVDA, GOOGL and AVGO near their October floors while calling the AI trade capital-constrained (“AI running out of cash, not chips”).

The fear is in other people’s gauges, not ours. Our FOM gauge sits at 48.8. The desks cite fear elsewhere — CNN’s gauge in fear for a month, hedge-fund shorts at a five-year high — none of which we hold on disk. The positioning that carries information is the option tape’s: the November fixed-strike hedging bid and an inverted index front across the Fed and the cloud reports.

Earnings — the AI beats were sold and the defensive beats were paid, and next week’s reports carry the same question

The label stays mixed; the split is by what the company sells. Our earnings regime reads the last large reports on their second day. The thirteen that completed day two on Thursday’s close read six beats paid (DHR, PM, ELV, ISRG, COF, CB), four beats muted (TMO, T, NFLX, SPCX) and three misses absorbed (ABT, MCO, UAL), none sold: mixed, so it adds nothing to our count. Thursday’s own reaction day leans one way inside that label. Four AI-linked beats were sold on day one — META, LRCX, TXN and NOK — while eight defensive, financial, defence, media and airline beats were paid (PG, SPOT, CMCSA, HON, LMT, NDAQ, AAL, and IBM back to flat after an after-hours drop). A market that sells strong reports from companies spending on, or selling into, AI while it pays strong reports elsewhere is pricing the spending, not the results. Friday’s close sets those labels; a run of AI beats sold on day two, with next week’s reports following, would turn the earnings regime into a bearish input for the AI groups.

After Thursday’s close, the same pattern. WDC reported record profit on storage pricing and guided its next quarter sharply higher; it traded 331.34 after the report and 345.52 overnight, 2.45% under its 354.19 close, inside the 320.44 to 387.94 range its options priced for Friday. BE reported revenue up 138% and raised its full-year guide, and traded only about 1% higher overnight at 219.90 after falling 8.70% on Thursday into the report, with negative quarterly cash flow named against it. RNG traded about 14.6% higher, DECK about 7.3% lower. NEM reported record adjusted earnings on gold sales and traded 1.4% lower with gold. These are after-hours prints and grade nothing until Friday’s close.

Next week the spending question goes to the spenders. The October 30 options, which hold the Fed and the PCE print as well as each report, price MSFT at about 5.6% each way (its median report move is 6.1%), GOOGL at 6.1% — far over its own 2.7% median — AMZN at 7.4% (median 5.9%), INTC at 11.1%, NOW at 10.4%, and SNDK, COIN and MSTR between 9% and 11%. Their October 30 implied vols run 1.30 to 1.48 times November’s, so the event premium sits in the front week. The options price how big the moves can be, never which way they go; what we will read is whether the market keeps selling AI spending when the cloud companies report it.

Timing — both readings of the October chart now draw the same low on Thursday or Friday, and they part across the Fed week

Thursday confirmed the high, and the low is Friday’s test. Savino’s October S&P chart comes in two versions, the original and its mirror image, and we grade both on every dated turn, as a clock and a sign — never as a price. Thursday’s lower close confirmed Wednesday 10/21 as a local high, the fourth dated turn in a row the original has landed at its fixed lead. Both versions now draw the same low on Thursday 10/22 or Friday 10/23 — the first time since the counts were reset on 10/16 that they put a turn on the same sessions with the same sign. A higher Friday close confirms Thursday as that low for both. A lower Friday close moves the candidate to Friday, and Monday’s close then decides; a lower Monday as well would be a miss for both.

After the low they disagree, which is why the clock says little about next week. Both draw a higher close on Monday 10/26. From Tuesday the original draws a rise into a high between 10/27 and 10/29 — across the Fed statement and the MSFT, GOOGL and NOW reports — and a fade into month-end; the mirror draws its high on Monday and a decline into a second low on 10/29 or 10/30. For Tuesday through Thursday the two signs are opposite, so the clock carries no net direction for the Fed week itself. Our own S&P projection, registered on 10/18, put the first low of this leg inside 10/22 to 10/29, and Thursday opened that window with its lowest close so far; Silva’s 2018 analog puts its marker for 2018’s first-leg low on Wednesday 10/28, the Fed day. Several clocks pointing at the same week are not several votes: two are one chart in two signs, and one is our own build.

The bond chart’s rise ran a third session. Savino’s 30-year bond forecast drew a drift higher in bond prices into a high around 10/28 to 10/29, and Thursday was its largest day: the 30-year future closed at 107.8125 and TLT at 81.74, over its weekly ceiling 81.36. If that drift runs into its drawn high, the levels in play are TLT’s zone ceiling 82.09 and its quarterly ceiling 83.27; magnitude always comes from the bands, never from the chart.

Scorecard — grading Wednesday’s edition on Thursday’s close (14.0 of 21 points, a B)

The frame held; the one-day labels did not. Our 10/21 edition called Wednesday’s rally a squeeze in last week’s losers, not a bottom, and kept every bearish group call standing through the one day they were tested against a broad AI bid; Thursday’s worst AI basket day on record answered both. Its Thursday sheet named the two levels the session turned on: the repair was undone under 7,713.88, and the zone floor 7,642.39 held on the close. Its option reads were tested on Thursday’s open-interest count, and 10 of 12 held (one at half weight).

What it got wrong. It suspended the AI groups’ price input for a session after a one-day repair, against its own sentence that a one-session repair needs a second close — Thursday restored the input. Two continuation labels on single instruments with one-day inputs (the bitcoin future on the stablecoin rule, T on its own report) ended the next session on a rotation day. NFLX’s “trend, not a bounce” read met a +3.09% day on which the market bought everything outside the AI trade. A META call spread written as bought turned out, on the count, not to be a new position. And both WATCH entries missed moves in their own direction: SMH fell 5.99% with its put spread not on, and TLT rose 1.87% with its call spread waiting at a floor it never revisited.

The grade. On the 21 lines that resolved with forecast credit — 11 hits at full weight, 5 at half weight, 1 partial, 4 misses — the edition scores 14.0 of 21, 67%: a B, against B-minus for the 10/20 edition and C-plus for 10/19. The list of structures earned a B-minus on its own: the index hedge paid on the down day, and every rule was followed as written; the weak point was the entry rules, one re-set a session late and one set where the bond never went. Our lesson for tonight’s list is the one we apply to the SMH put spread below: re-price a WATCH at every close, and say what would make it stale.

Friday and next week — rebound or not: the overnight tape, the conditions and the levels

Our read for Friday: the index carries a measured rebound setup, the overnight tape leans the same way, and a bounce in the AI complex is a counter-trend move inside a bearish group trend, not a bottom. We do not publish path weights; here is the evidence on each side and the lines that decide it.

For a Friday rebound. (1) The index cell: a close under the daily floor in the current tightening regime has come back inside the next session 6 times of 8 since the 9/16 hike (a small sample measured on the indices only), and the S&P closed under both its daily floors with every swing band intact and 173 points over the weekly floor. (2) The overnight tape (01:37 Central, context only): S&P futures flat at 7,702.75 against 7,706.25 at Thursday’s cash close, Nasdaq futures up 0.2% after an overnight low, the AI names bid — SMH 531.60 (+1.6%), MU 880.28 (+2.8%), SNDK +2.7%, AMD +1.8% — while staples and bonds gave back a little. (3) Korea: the KOSPI opened at a six-month low on its Friday session, -4.1%, and closed up 0.3% after SK Hynix pulled its Solidigm US listing — the memory market reversed overnight. (4) The clock: both readings of Savino’s October chart draw a low on 10/22 or 10/23 and a higher close to confirm it, and our own projection put the first low of this leg inside 10/22 to 10/29. (5) Credit stayed calm, so the selling has not become forced. (6) Friday is a weekly expiry: the protection bought into Thursday’s fall rolls off, and the hedging bid in the November strikes is already in the price.

Against it. (1) All 19 of our bearish group calls regained their price input on Thursday, the AI basket had its worst day on record, and the group’s lower-high, lower-low trend is intact. (2) The S&P closed 48.57 points under Silva’s gamma flip (7,712), the side where dealer hedging adds to moves, and the reversion-buy gate is closed (VIX over its zone middle, the index under the flip): a low with that gate closed is not a buy signal. (3) Crude traded 100.92 overnight and Brent 114.60, through the 100 line the Maverick ties to the Nasdaq. (4) Next week carries the Fed (Wednesday) and the megacap prints that set Thursday’s tone — MSFT, GOOGL and NOW after Wednesday’s close, AMZN and INTC after Thursday’s — on the same spending question META raised. (5) The rates-at-the-record window that opened on 10/5 runs to 11/2, and its measured median worst close (7,590.48) sits 73 points under Thursday’s close.

The Friday conditions.

• Rebound confirmed: an S&P close over 7,663.43 grades the rebound setup a hit and confirms the low both chart readings draw; a close back over the 7,712 flip and Friday’s daily ceiling 7,722.12 (zone middle 7,728.09) is a real repair, which puts the week’s ceiling 7,828.43 back in play for next week. For QQQ, a close back over 714.96 (Friday’s daily ceiling) and 717.90 (Thursday’s floor).
• Rebound failed: an S&P close under the zone floor 7,629.03 and Friday’s daily floor 7,604.74 makes next week a test of the week’s floor 7,490.45; a QQQ close under its October floor 702.09 is a monthly-band break and the exit line on the operator’s QQQ call.
• The AI complex: SMH back over 556.19 repairs its October break, and the 556 to 568 zone is where our rebound map buys bearish structures; an SMH close under 514.05 (its 200-day) puts the group in a trend below its 200-day, and under 504.22 (the July low) confirms a lower low on the fund itself.
• The rotation: RSP against QQQ. Staples over their October 2-sigma ceiling (86.86) carry no measured basis for a fade, so the defensive side is not where the bounce trade lives.
• The prints: the flash PMIs at 8:45 Central (manufacturing forecast 55.5 against 55.9, services 58.0 against 58.8) and the final Michigan sentiment at 9:00 (forecast 46.3 against 48.1).

Next week, day by day (Central; weekdays computed). Monday 10/26: VZ before the open, the 2-year auction; the inverted reading of the Savino chart draws its high here. Tuesday 10/27: durable goods (forecast +0.6%), consumer confidence, the 5-year auction; KO, UPS, BA, HCA, FCX, PYPL, SOFI before the open, V, STX, NXPI, HOOD after. Wednesday 10/28: the Fed at 13:00 (the consensus a hold), the 2018 analog’s session-16 marker; MSFT, GOOGL, GOOG, NOW after the close. Thursday 10/29: third-quarter GDP and core PCE (forecast 0.3%) at 7:30; CAT, LLY, MA before the open; AMZN, INTC, MSTR, COIN after. Friday 10/30: month-end, the employment cost index, the crypto expiry cluster; XOM, CVX, ABBV before the open. The two chart readings agree on a high early in the week and part after it: the inverted one turns down from Monday into a second low on 10/29 or 10/30, the original holds up into 10/27 to 10/29 and fades after the Fed. Our registered AI rebound map expects a rebound toward SMH’s rebound zones into the 11/3 midterms, then a second leg lower after the December expiry. The week’s magnitude comes from the bands, not the charts: next week’s Silva weekly board posts after Friday’s close.

Bottom line

The repair is undone, and the money rotated rather than left. Company news on AI spending — META’s budget and the stock class to fund it, the NVIDIA chief on how the new cloud builders pay for chips — sold the AI hardware chain, power equipment, the miners and the speculative names, and gave our AI basket its worst day on record. The same money bought bonds, staples, health care, regulated utilities, banks and the equal-weight index. Credit did not follow the equities down, and none of the five lines we registered for a liquidity drain fired. The S&P closed under its daily floors and inside every swing band.

We call no index direction; the calls stay on the groups. On the AI groups our count has two bearish inputs (the rate level as a standing condition and the price trend, restored by the failed repair), one short of a full call, and the 19 bearish group calls and three registered sector views carry the lean with their retire lines. For Friday, the index carries a measured rebound setup — a close over 7,663.43 — and the overnight tape leans that way; a bounce in the AI names is a counter-trend move inside their trend, and our registered map buys bearish structures in the rebound zones, not at the low.

What would change our read: an S&P close back over 7,712 and 7,722.12 repairs the index again and puts the week’s ceiling in play for next week; a close under 7,629.03 and 7,604.74 opens the week’s floor 7,490.45. For the AI groups, SMH back over 556.19 ends the day’s break and enters the rebound zone; SMH under 514.05 (its 200-day) and 504.22 (the July low) confirms the group’s lower low on the fund itself. A rise in yields, HYG under 76.39 or two of our liquidity-drain lines firing would turn a rotation into a drain. Next week’s Fed and the cloud companies’ reports answer the same spending question META raised.

Top Trades to Follow

This is the list of institutional-style option structures the next edition grades. It is a record for grading, never advice and never a position size. Five structures carry over, two stay at WATCH, and one is added. Two lines printed on Thursday and each is handled by its own rule: the SMH put spread’s entry condition printed far below the zone it was built for and is declined at the low (below), and the TLT spread’s entry level did not print. Marks are our estimates for Thursday’s close, from the day’s last option trades moved to the closing price, not closing quotes. Every entry states where the name’s long-dated vol sits in its own history, the level, the exit, and its maximum loss in dollars. On Thursday’s file 36 of 49 measurable names carried a front month at or above 1.10 times their back month, so the list keeps three term structures. Our entry check is a test of entry quality, not a forecast: on its own record, names that passed it did no better than names that did not.

Running record: about -506 dollars on fourteen entries since 9/28 at these estimated marks — nine closed at about -605, the five carried about +99 (SPY +76, XLU +61, EWZ +8, NVO +5, AMZN -51); the SMH calendar is the fifteenth.

Sources

Expected moves: Silva’s daily expected-move boards and zone sets for 10/22 and 10/23 (board image, zone document, zone visual, zones image, range-and-trend image and the TradingView data file), his weekly board for 10/19 to 10/23, his October monthly board and his fourth-quarter board with the year-end collar levels, the FOM sentiment gauge for 10/22 and our sentiment tracker, and our option-implied band ledger computed from Thursday’s options file. Tradytics: the 10/22 options and dark-pool files (16:15 Central) and the 10/21 files for day-over-day comparisons, the 10/22 options dashboard (all 24 pages read as images), and our flow, dark-pool, monetization, setup and long-dated vol ledgers. Recon census: all 651 per-ticker files for 10/22, the sector chunks and summary, the per-ticker dark-pool captures, the name-behavior ledger and its roster, the census screen, the AI and speculative basket monitors, the hourly cyber-and-software top watch and memory-and-semis bottom watch, and the group-convergence state. Intraday: the 10/22 close recap, the four cut briefs and the pre-open brief, the level frames, the scheduled-prints files, the news and sector sweep files, the FOM and Maverick morning briefs with Silva’s gamma anchors, the analyst feed and the InvestAnswers posts, and the 10/22 chat session on rebound trades, the AI-basket view and the liquidity watch lines. Timing: Savino’s October S&P projection in both versions with the 10/19 updates, his 30-year bond forecast in both versions, our own S&P projection and AI rebound map, the Silva 2018 analog tracker and our timing ledger. Commentary: the polished 10/22 transcripts from Arete Trading, Benjamin Cowen, Darius Dale (42 Macro), Dividend Talks, ET Tradytics, FX Evolution, Financial Freedom 101, Geeks of Finance, Ilya Spivak, InvestAnswers, Krown, the Maverick, Matt Hougan, Mike Jones, Rob’s Child, SpotGamma, Tim Knight, Tyler S and tastylive; Mike Silva’s 10/21 deck and video (no Silva file dated 10/22 was on the shelf). External data: FRED (the high-yield spread, the 10-year real yield, the breakeven and the 10-year constant-maturity yield), TradingView daily and hourly bars (the S&P 500 and its averages, the 10-year, S&P, Nasdaq and Russell futures, crude, Brent, gold, the dollar index, VIX futures, the Cboe one- and three-month implied correlation indexes, bitcoin, the KOSPI and the Nikkei), the TradingView economic calendar and news headlines, and broker after-hours quotes as context. Prior cycle: the 10/21 comprehensive analysis, the 10/21 daily report and its regime snapshot.