US Market Recap | 2026-09-09 (ET) Wednesday

Coverage window: 2026-09-09 09:30 ET open → 16:00 ET close; the after-hours section is read through 17:08 ET. The prior session was 9/8 (Tuesday); "previous close" throughout refers to the 9/8 close.

Price conventions (every figure is tagged with its read timestamp — do not compare across timestamps):

• Open/high/low/close, volume, volume ratio (vs 10-day average volume) and after-hours prices for single names and ETFs come from the CNBC quote API; regular-session figures are read at 16:00–16:15 ET, after-hours figures at 17:08 ET.
• Treasury yields: direction and levels use the official Treasury yield curve CSV (15:30 ET fixing); intraday extremes use CNBC real-time. The two are not from the same source; a few bp of difference is not read as directional. ⚠️ CNBC's Treasury change_pct field was again self-contradictory this time (the 2Y last is above CNBC's own previous close yet the field shows a negative change). This report does not rely on that field at all, and instead computes everything from the difference between the two daily fixings.
• Crude, refined products, gold, natural gas and equity index futures come from CNBC, read at 16:50 ET, and are the last electronic-session trade, not the NYMEX 14:30 ET settlement price.
• Crack spreads are computed locally: same-month contracts, ×42-gallon conversion, 3-2-1 weighting; both legs use the 16:50 ET read and the 9/8 previous close, so they are same-source comparable.
• Market breadth is computed locally: 501/503 S&P 500 constituents (BF-B and BRK-B failed on ticker format), each computed from the CNBC close versus previous close. This is S&P 500 breadth, not NYSE/Nasdaq all-market advance-decline counts — the latter was not obtained this time, so please do not conflate the two.
• Company financials come from primary SEC 8-K filings and company press releases, tagged at each point of use.

⚠️ This report's reconciliation tables carry a "close vs pre-market price" column. Reconciling on the daily percentage change alone systematically inflates the hit rate: this morning's crash names had already finished falling before the open, and announcing after the close that "we told you to avoid it" is not honest. That column is the real P&L of executing the list as written.

---

0. One-Line Recap

Today's real driver was not on any line of the pre-market list — it was rates. And the irony is that the pre-market list wrote it up as scenario ④ under "conditions that would invalidate this report's conclusions," yet never treated it as a theme.

1. Risk-off — and risk-off with terrible breadth. S&P 7,636.36 (−0.48%), Nasdaq 26,253.34 (−0.64%), Dow 52,380.66 (−0.77%), Russell 2000 2,921.24 (−1.32%); the three major indices fell for a third consecutive day. But the index declines badly understate the internal damage: S&P 500 constituents were 98 up / 403 down, an advancing share of just 19.6%, with a median of −0.89%; only 126/501 (25.1%) rose from open to close — the index opened at the high of the day (7,660.68) and ground lower from there.

2. The driver was disappointment over the Treasury buyback operation. Treasury announced it was raising long-end liquidity-support buybacks to $6 billion (three times the $2 billion originally communicated), but the market judged it "not enough": the 10-year yield spiked to 4.857% intraday (CNBC), with an official fixing of 4.83% — the highest fixing in the 173 trading sessions of 2026. 2Y 4.43% (+4bp), 30Y 5.28% (+3bp) — essentially a parallel bear shift. (Treasury press release¹ · CNBC²)

3. The rate attribution is confirmed three separate ways in the sector structure — it is not a story constructed after the fact: 0 of 31 utilities constituents rose and only 1 of 30 real estate names rose — the two longest-duration sectors were almost completely wiped out; Russell 2000 −1.32%, the worst of the four major indices; and software (IGV −0.81%) fell while semis (SMH +0.10%) did not. ⚠️ Trigger ④ written in §8.5 of the pre-market list read "10Y decisively breaks above 4.85% → the software bounce is the first thing to be interrupted" — that fired precisely today, but it was filed under "risk disclosures" and never made it into the theme ranking.

4. The strongest theme was crude, but the equities barely followed. WTI $96.74 (+3.99%), Brent $101.66 (+3.82%) (16:50 ET), yet XLE rose only 0.83%; just 12 of 21 energy constituents advanced, with an equal-weighted mean of +0.29%. The pre-market criterion "if XLE's gain ÷ WTI's gain < 0.4, upside is limited" — today that ratio was 0.21, deteriorating further from yesterday's 0.37; the criterion held and the direction was called correctly.

5. The single biggest stock event of the day had nothing to do with the list: META +6.55%, on the release of its standalone AI agent Muse; GOOGL fell 2.28% the same day — that mirrored up/down pair is better evidence that this is a product narrative rather than a broad market rally than the size of META's own gain.

6. The honest conclusion from reconciling the pre-market list: 3/4 up on closing prices looks decent, but only 2/4 by each name's own written verification point, and executing at the pre-market price was a loss (mean −0.85% vs SPY −0.46%). See §2.

In one line: crude breaking $100 was the loudest headline of the day but only the second most important thing; what actually drove 403 stocks lower was the 10-year yield making a new high for the year — and the pre-market list put that in the disclaimer, not in the theme table.

---

1. Market Overview

1.1 Indices (CNBC, closing basis)

All three major indices fell for a third consecutive day.

1.2 Market Breadth (computed locally, 501 S&P 500 constituents)

⚠️ The divergence between the index and breadth has to be spelled out: the S&P was −0.48%, while the equal-weighted mean was −0.96% and the median −0.89% — the index fell only half as much as the median stock, and nearly the entire gap is contributed by a single name, META (+6.55%). Saying "the market only fell 0.5%" would badly misrepresent what holding positions actually felt like today.

The most striking single case: only 1 of the 24 communication services constituents rose (that being META), with an equal-weighted mean of −1.88%, the worst in the market; yet cap-weighted XLC fell only 0.62%. Same sector, two yardsticks 1.26pp apart.

1.3 Sectors (left: S&P 500 GICS equal-weighted, computed locally; right: sector ETF, cap-weighted)

Other ETFs: SMH +0.10%, IGV −0.81%, ITA −1.84%, XOP +0.57%, OIH +0.18%, KRE −1.16%, XBI −1.57%, ARKK −1.77%, TLT −0.57%, HYG −0.18%.

Only two sectors did not fall — energy and information technology — and within information technology only 28 of 73 names rose. "Tech held up" is an illusion created by cap weighting.

1.4 Rates, Volatility, Dollar, Commodities

⚠️ The dollar index is suspect and has been cross-checked: the .DXY close CNBC gives, 98.787, is identical to the previous close to three decimal places, while the day's range was 98.599–98.90 — a character-for-character match is unlikely, so the first suspicion is a stale field. We cross-checked locally using the dollar ETF UUP as a proxy: UUP −0.04%, volume 1.37M (1.10× volume ratio), so it was indeed roughly flat. ⇒ We accept the qualitative call that "the dollar was roughly flat on the day," but we do not use the absolute .DXY figures for any fine-grained comparison.

1.5 Sentiment Read

Risk-off, but not the panic kind — the discount-rate kind.

• VIX only rose to 16.46, still low; HYG was just −0.18%, credit did not weaken alongside equities ⇒ this is not a credit event.
• But TLT −0.57%, XLU 0/31, XLRE 1/30, IWM −1.32% ⇒ the pressure landed precisely on long-duration assets.
• XLK closed flat while IGV fell 0.81%: both are "tech," yet high-multiple software fell while low-multiple hardware and semis did not. ⚠️ This is exactly the fork that could have falsified the rate attribution — and today it points to "the rate attribution holds": low-multiple, near-dated-cash-flow legacy tech (IBM +3.38%, HPE +5.12%, HPQ +4.92%, AKAM +4.87%, FFIV +3.85%) rallied as a group, while high-multiple software (NOW −2.31%, ADSK −2.63%, CRM −1.99%) weakened as a group. ⚠️ A limitation that must be disclosed alongside it: we could not locate a 9/9 primary catalyst for HPE / HPQ / IBM / AKAM / FFIV individually. A third party attributed HPQ's gain to "dividend analysis," but HPQ's ex-dividend date was precisely 9/9, and going ex-dividend can only depress the price, not lift it; that attribution fails mechanically and this report does not accept it. "Low-multiple rotation" is a description of price action, not a conclusion supported by a primary catalyst.

---

2. Pre-Market List Reconciliation

2.1 §9① The 4 Names Most Worth Watching Today

Pre-market price = 9/8 close × (1 + pre-market change), where the pre-market change is the 08:50–08:55 ET read as printed in the list.

Hit rate (three bases; all three must be viewed together):

A one-line self-critique: names ①, ②, and ③ on the list are essentially the same oil trade, and the pre-market note itself disclosed this concentration flaw (§9① footnote 2) — but disclosing it is not the same as fixing it. Today the commodity leg was entirely right (WTI +3.99%) while the equity leg only half delivered; of four positions, only XOM's verification point passed cleanly. What is more worth remembering is COP: its verification point was set precisely as "can it outperform VLO," and today VLO (+1.59%) outgained COP (+1.10%) by 0.49pp — the criterion failed by a small but unambiguous margin, and it cannot be waved away on the grounds that "upstream still beat refining overall."

⚠️ The "upstream > refining" ranking deserves a fair record too: although COP individually lost to VLO, the 5 refiners were +0.36% equal-weighted, still trailing XLE (+0.83%) and the integrateds (XOM +2.22%, CVX +1.91%). The ranking holds on group means and fails on the specific pair that was singled out as the criterion. The criterion was self-imposed; losing is losing.

2.2 §6 Negative-Catalyst / Avoid List

Avoid-list hit rate (two bases, enormously different):

This is exactly why the "close vs pre-market price" column has to be there: the vast majority of that −14.73% occurred before 09:30 and was already in the price when the pre-market list was published. The list's genuine incremental value was only TTAN (−13.63%) and BRZE (−11.46%) — those two were real calls that kept delivering after the open; the other five essentially "reported something that had already happened."

2.3 Other Directional Conclusions

2.4 Three Gaps Left Open Pre-Market; Two Are Closed Today

---

3. Theme Verification

3.1 Themes Missed Entirely Pre-Market (in descending order of importance)

3.2 VRT −9.61%: An Attribution That Must Be Left Blank

VRT was the second-largest decliner in the S&P 500 on the day (behind only CASY), at a 2.43× volume ratio with ample turnover — this is not thin-volume noise.

Third parties broadly explain it as "profit-taking in the AI data center infrastructure group." We tested that claim directly against a peer group:

⇒ VRT fell 6.3pp more than even the worst peer, and every peer's volume ratio was ≤1.0 (i.e. no unusual volume), while only VRT was at 2.43×. "Sector profit-taking" cannot explain this distribution. ⇒ We did not find a 9/9 primary catalyst for VRT (no 8-K, no rating change, no company announcement). This report labels it "attribution unknown" and does not speculate. This is listed as a gap that must be closed next session.

---

4. After-Hours Earnings Moves (after the 9/9 close)

⚠️ Read at 17:08 ET, with volume tagged for every name. After-hours quotes are cumulative, so an after-hours price on thin volume is not a fact; and an after-hours price before versus after the conference call are two different things, noted separately below.

---

5. Flows and Sentiment

5.1 The Direction of Sector Rotation: Not "Defensive vs Offensive" but "Short Duration vs Long Duration"

Leaders (only two sectors positive): energy +0.29% (12/21), information technology +0.05% (28/73). Laggards: communication services −1.88% (1/24), consumer discretionary −1.60%, consumer staples −1.51%, industrials −1.41%.

⚠️ The most noteworthy structural feature today: the traditional "defensive rotation" did not happen.

• Utilities −1.15%, 0/31 advancing — on a risk-off day, the most classic defensive sector had among the fewest advancers in the entire market.
• Consumer staples −1.51%, 5/33; health care −0.56%, 13/59 — none of the three defensive sectors was positive.
• ⇒ If today were "flight to safety," money should have gone into utilities and staples; instead they were among the hardest hit. ⇒ Today was not a flight to safety, it was a rise in the discount rate, and utilities / real estate / small caps carry the longest duration.

This is an independent corroboration of the IGV/SMH fork in §1.5: two paths (sector-internal structure and tech-internal structure) point to the same attribution, and neither depends on interpreting the news — only on prices.

5.2 VIX and Treasuries

• VIX 16.46 (+4.71%), open 15.65, high 16.68. The absolute level is still low — this was an orderly repricing, not a panic.
• HYG −0.18%: high-yield credit barely moved ⇒ the pressure is not in credit, it is in rates.
• TLT −0.57%, consistent with the 10Y +3bp.
• 2s10s narrowed from 41bp to 40bp: essentially a parallel bear shift, not a meaningful bear flattening. ⚠️ A 1bp move is within the noise of day-to-day fixings, and this report draws no directional read on curve shape from it (last year there was an episode in which relying on CNBC's broken Treasury change field wrote the entire "bear flattening = rate-hike repricing" line backwards; this time we switched to computing it ourselves from the Treasury CSV).

5.3 The Commodity Leg: Oil Is Rising, Refining Margins Are Compressing — the Pre-Market Quantitative Pivot Still Holds on a Closing Basis

Computed locally (CNBC same-month contracts, 16:50 ET read vs the 9/8 previous close, same-source comparable):

The 08:33 ET pre-market read was 3-2-1 (Brent) −1.43%; on a closing basis it is −1.86% — not only no reversal, but further deterioration. The two qualifications the pre-market note placed on that figure should be recorded alongside it: ① it genuinely was only a snapshot at the time (it had swung from +0.14% to −1.43% within 36 minutes), and the closing basis proves that caution was warranted; ② the absolute level of crack spreads remains extremely high (3-2-1 around $61/bbl, distillate crack above $100/bbl), so refiners are making a great deal of money right now. This report's argument has always been "it did not get better today," not "refiners are not making money." ⚠️ At the same time it must be conceded: this correct commodity logic did not translate into a correct equity ranking today. VLO still rose 1.59% and beat COP.

5.4 Risk-On / Risk-Off Characterization

Discount-rate risk-off (day 3). Three independent pieces of evidence: ① breadth of 19.6% (dreadful) while VIX was only 16.46 (no panic); ② credit (HYG −0.18%) did not follow; ③ the damage was precisely concentrated in long-duration assets (utilities 0/31, real estate 1/30, Russell −1.32%, high-multiple software −1.56%) rather than in high beta (semis, SMH +0.10%).

---

6. Next-Session Outlook (Thursday, 2026-09-10)

① Theme Continuation

② Tomorrow's Earnings and Macro Calendar

⚠️ ORCL basis warning (carried over from the pre-market list and restated): the widely cited FY27Q1 consensus EPS of $1.299 and the company's own non-GAAP guidance of $1.72–$1.76 differ by 26%, and cannot be on the same basis (the former is very likely GAAP). If you see a headline tomorrow after the close reading "a massive 30%+ beat," the first thing to do is check the basis, not add to the position. Also, the revenue consensus of $19.13 billion is exactly the midpoint of the guidance range — it is an echo of the guidance, not an independent forecast. (The consensus is a third-party aggregate; the contributing institutions and cutoff date were not obtained.)

③ Key Names to Watch (Ticker + Verification Point)

Design constraint (self-imposed here): today's lesson is that "a watchlist with only one dimension" — 3 of the 4 pre-market names were the same oil trade. The 5 lines below are deliberately tied to 5 different driving variables, and every verification point tests "the thing most likely to make this call lose money" rather than "whether it will go up."

⚠️ Position constraint (based on today's specific lesson): lines 1, 3 and 4 are all single-company events with unknown persistence, and together should not exceed half of total watch capacity. UNH's failure today is exactly the case in point: the reason given pre-market was "the only positive 5-day return," a purely positional metric — a name with no primary catalyst should not occupy a priority slot.

④ What to Avoid

1. The first tier of AI power / neoclouds (VRT / CRWV / NBIS / SMR / NNE / WULF / APLD) — 0/6 advancing, equal-weighted −4.91%; CRWV and NBIS both gapped up and faded, and the distribution criterion has fired.
2. Directional positions in high-multiple application software — the bounce died the same day (1/7), and ADBE carries its own event risk after tomorrow's close; buying options now means taking on both direction and IV crush.
3. Reflexive buying of "beats," but with grading — ⚠️ today's 4 cases were 2 up and 2 down (CHWY −10.83%, COO −11.43% after hours vs ODD +26.48%, GME +5.29%); the pre-market pattern of "the 4 cases found were all down" has been broken today. The correct rule is not "a beat means a decline," but: first ask whether it beat consensus or the company's own prior guidance, then ask whether cash flow and leading indicators beat as well — CHWY beat on both revenue and EBITDA, and the reason it fell was free cash flow −15.5%.
4. Utilities and real estate — 0/31 and 1/30; in a week when the probability of an FOMC hike is above 50%, the "defensive" label has already stopped working today.
5. Broadband / cable (CHTR / CMCSA / TMUS) — sustained high-volume multiple compression with no primary catalyst; a decline with unknown attribution is not to be traded, on either side.
6. VRT — same reasoning as above; an unexplained crash is as untradable as an unexplained spike, on either side (consistent with the pre-market treatment of ODD; ⚠️ and ODD demonstrated the cost of this discipline today: its catalyst was in fact public at 07:00 ET and we missed it in our scan. The discipline is not wrong; the search depth was insufficient).
7. After-hours quotes with no volume behind them — today's after-hours prices for RH (32,744 shares), OXM (14,967 shares) and the like are not facts.

⑤ Inputs for Tomorrow's Pre-Market List

1. Rates must be fixed first, themes second. Today's lesson is that rates were written into the disclaimer. Before the 08:30 PPI print tomorrow, no ranking of duration assets holds; the first section of the list should be "where the 10Y sits before and after PPI," not some sector.
2. Do not treat "AI infrastructure" as one theme. Power/neoclouds and optical/interconnect were 5.4pp apart today. Split it into at least three legs: power and neoclouds; optical and AI interconnect; front-end equipment and memory.
3. Pre-market single-name breadth cannot be extrapolated. This morning's "semis 16/16 down" was the most misleading reading of the day — SMH closed +0.10%, and MU/MRVL/ALAB/AMD all rallied hard. Pre-market breadth describes only the pre-market.
4. The reconciliation table must keep the "close vs pre-market price" column. Today's avoid list was −14.73% on closing prices and +1.47% on pre-market prices; those two numbers tell two completely different stories.
5. Add a scan of the 07:00–08:00 ET pre-market earnings window. ODD's catalyst was public at 07:00 ET while the list's read timestamp was 08:50 — that window directly produced one mislabeled "attribution unknown" today.
6. Add a "next 5 trading days calendar" table. We missed the pricing of the 9/16 FOMC hike today because the news scan only looks for "new news," and the FOMC is a calendar event, not news.
7. Close three gaps: VRT's primary catalyst, the cause of the defense sector's decline, and SSL's FY2027 hedge ratio and strike prices (SSL was +9.62% today, the strongest name in the "watch only" bucket, and that variable is still blank).

---

Ops Notes (not sent to clients)

Data collection and tooling status this run:

1. yfinance was not used. The slot prompt says "WebSearch + yfinance," but per the memory record our yfinance has long been rate-limited by Yahoo (HTTP 429); we skipped it outright this time with no exploratory calls (to avoid the old trap of "hammering our own IP with concurrency"). All quotes were switched to CNBC quote.cnbc.com/quote-html-webservice with exthrs=1, which returns OHLC + volume + tendayavgvol + pcttendayvol + ExtendedMktQuote after-hours price/volume in one call. During the after-hours session that endpoint's fields are complete, consistent with the memory record "CNBC extended-hours field asymmetry" (available after the close, unavailable pre-market). Recommend locking this in as the default channel for the US recap slot.
2. CNBC's Treasury change_pct field was broken again and was caught. US2Y returned last 4.434% / previous_day_closing 4.398% (i.e. yields rising) yet gave change_pct −0.0703%. Relying on it would have written "the 2Y fell" into the curve shape and inverted the entire rate-hike pricing conclusion. We switched to computing from the two daily official Treasury CSV fixings and told clients in the conventions box in the body that we do not rely on that field. This is the second hit for memory cnbc-treasury-changepct-field-broken; recommend blocking that field at the data-fetch script level.
3. .DXY looks like a stale field; rescued with an ETF proxy. CNBC gave last == previous_day_closing == 98.787 (identical to three decimals) with change_pct = UNCH. Per memory cboe-stale-symbol-silent-trap, fields that "look like normal values" are the most dangerous. We cross-checked with UUP (−0.04%, 1.37 million shares, 1.10× volume ratio), confirmed "the dollar really was roughly flat," and so kept the qualitative call while discarding the absolute figures. This run was a positive application of memory etf-proxy-detects-stale-spot-quotes.
4. WebFetch was blocked in two places: CNBC's closing live blog returned 403 and Yahoo Finance's closing live blog 429. As a result the day's market narrative came entirely from WebSearch summaries plus locally computed data, with no primary live-blog source. The impact is contained — every number that entered a conclusion was computed locally, and search summaries were used only to locate events (the Treasury buyback, META Muse, the CHWY/COO/ODD earnings).
5. ⚠️ The biggest trap caught this run: when third-party summaries conflict with locally computed data, the local computation always wins.
   • A Motley Fool intraday piece on 9/9 wrote "energy and utilities gained." Our own computation: 0 of the 31 utilities constituents rose, equal-weighted −1.15%, XLU −1.17%. Copying that would have inverted the entire "discount-rate risk-off" line in §5 into "defensive rotation."
   • Alphastreet wrote that HPQ "closed at $32.70, +5.9%"; our CNBC read is $32.39 / +4.92%; and it attributed the gain to "dividend analysis," while 9/9 was precisely HPQ's ex-dividend date, and going ex-dividend only depresses the price. Neither claim was accepted; HPQ is tagged attribution unknown.
   • One third-party summary put AVAV's non-GAAP EPS at $0.75, while SEC 8-K EX-99.2 shows $0.59. We used the primary filing and disclosed in the body that the discrepancy is unresolved.
   • Quiverquant's "Why VRT Is Down Today" was actually published on 8/18 (its content is −6.6% and a Q2 revenue miss). Memory recurring-column-headlines-hide-date was a hit; not accepted.
   • One summary claimed HPE's gain was based on "revenue +34% and raised 26/27 guidance" — that is old news from the 9/3 earnings report, not a 9/9 catalyst. Likewise not accepted.
6. Reverse-computation check on CASY's previous close. CNBC gave CASY change_pct −14.24%; reverse-computing previous_day_closing gives 733.49, consistent with the 9/8 close recorded in the pre-market list ⇒ pass (memory price-triple-must-be-reverse-computed). OCC's change_pct returned UNCH, which momentarily looked like stale data, but open 12.19 / high 14.10 / low 11.68 / 1.07 million shares (7.89× volume ratio) prove there was a full day of trading and it really did close at the same price as the previous close ⇒ a genuine round trip, not a field failure. This one was nearly mislabeled as a data-collection failure.
7. Breadth was computed successfully but with a limited basis. The S&P 500 constituent list came from a GitHub datasets CSV (503 names), and each was pulled from CNBC quotes, with 501/503 succeeding (BF-B and BRK-B failed on ticker format — the same two as in the pre-market list). ⚠️ This is S&P 500 breadth, not NYSE/Nasdaq all-market advance-decline counts; the latter was not obtained this time, and the conventions box in the body explicitly tells clients "do not conflate the two."
8. Sub-agents were not enabled, per project constraints. Historically in this slot fundamentals-analyst / risk-auditor have been the main source of corrections (see the ops notes for the 9/9 pre-market run: they overturned 1 theme characterization, 1 recommended name, and 3 factual errors). The system constraints for this session explicitly required not calling the Agent tool, so all verification was done independently by the main process. ⇒ This means this report lacks an independent adversarial QC pass and is less reliable than past editions that had an auditor configured. Recorded here as-is; not reflected in the client version.

Three methodological problems discovered about ourselves this run (more worth recording than the data problems):

1. A correct judgment was filed in the wrong place. Item ④ in pre-market §8.5 ("10Y decisively breaks above 4.85% → the software bounce is the first thing to be interrupted") was fulfilled word for word today, yet it was written under "conditions that would invalidate this report's conclusions" rather than in the theme table. The result: we correctly predicted the day's main driver but allocated nothing on the back of it, and instead gave three slots to the same oil trade. ⇒ "If something in the risk disclosures gets validated, it is no longer a risk — it is a theme." Recommend adding a step after §8.5 next edition: go through each item and ask "if this happens today, how does the theme table have to change?"
2. The verification points were set correctly, but the conclusions were not constrained by them. COP's verification point ("if it cannot outgain VLO, the core ranking is meaningless") clearly failed today, while XOM's and OXY's criteria barely passed. If the pre-market note had treated "COP must beat VLO" as a real constraint, COP should not have taken a second slot within the energy group. This is another hit for memory verification-points-must-constrain-conclusions.
3. Missing ODD was a search-window problem, not a judgment problem. The catalyst (Q2 earnings plus raised guidance) was publicly released at 07:00 ET, while the pre-market list's read window was 08:50–08:55, so the news scan never covered the 07:00–08:00 stretch. "Attribution unknown" was therefore a false positive. ⇒ Recommend adding a fixed 07:00–08:30 ET pre-market earnings window scan to the US pre-market slot.

Gaps carried into the next edition:

• ① No primary catalyst found for VRT −9.61% (2.43× volume, 6.3pp worse than peers) — the body has already falsified the third-party "sector profit-taking" explanation using a peer comparison, and set a locked-in criterion.
• ② No primary catalyst found for CHTR −8.13% / CMCSA −6.61% (both on elevated volume) — third parties explicitly wrote "no fresh filings or analyst actions," but we did not independently check SEC filings or the two companies' IR pages either.
• ③ No primary catalyst found for the collective rally in HPE / HPQ / IBM / AKAM / FFIV — at present all we have is the price-action description "low-multiple rotation."
• ④ Cause of the defense sector's decline (carried over from gap ① of the 9/9 pre-market list; a second consecutive edition without closure).
• ⑤ SSL's FY2027 crude hedge ratio and strike prices (carried over from pre-market gap ⑤) — SSL was +9.62% today on a 3.07× volume ratio, the strongest name in the "watch only" bucket, and the only variable that determines whether it has fundamental support is still blank. A second consecutive edition without closure; priority should be raised.
• ⑥ COO's full conference call Q&A not obtained — the −11.43% after-hours reading falls within the call window and must be re-checked tomorrow morning.
• ⑦ The $0.59 vs $0.75 discrepancy in AVAV's non-GAAP EPS is unresolved — this requires reading the body of the EX-99.1 press release (we only read the EX-99.2 presentation materials this time).
• ⑧ NYSE / Nasdaq all-market advance-decline counts not obtained — this report's breadth is on an S&P 500 basis only. If the next edition wants to draw conclusions about "small-cap breadth," this must be added.
• ⑨ The primary clinical data text for TYRA / PYXS is still not obtained (carried over from pre-market gap ③). TYRA was +16.21% open→close today, and the pre-market decision to "take no position in either direction" saved 22%, yet to this day we still do not know whether that data was good or bad.

---

⚠️ Risk disclosure: this recap is a post-close summary of information and observations only and does not constitute investment advice. Regular-session data in this report is read at 16:00–16:15 ET, after-hours data at 17:08 ET, and futures at 16:50 ET as the last electronic-session trade (not a settlement price); after-hours quotes continue to change as conference calls progress and may be invalidated at any point before the next open. Consensus estimates and third-party data cited herein are quotations, not the views of this report, and do not constitute price targets. Market breadth is computed locally from S&P 500 constituents and is not an all-market basis. US equities carry high volatility and after-hours gap risk, and post-earnings IV crush and guidance reversals occur; automatically generated content may contain stale information or factual errors — company disclosures and SEC filings prevail, and this may not be used directly as a basis for trading.