U.S. Market Recap | 2026-08-05 (ET) Wednesday

Coverage window: 2026-08-05 09:30–16:00 ET regular session + after-hours through 18:05 ET. Quote snapshot taken at 18:05 ET, sourced from the CNBC real-time quote API (quote.cnbc.com); regular-session figures are the official 16:00 ET close, after-hours figures are the latest POST_MKT trade. Macro data sourced from official ADP and ISM releases. All times are U.S. Eastern Time (ET).

---

0. One-Paragraph Recap

1. Today was neither a broad rally nor a broad selloff — it was an internal rotation of "defensives + precious metals vs. growth + semiconductors": the Dow +0.49% to 54,349.12, its third consecutive record close; while the Nasdaq was -0.83%, the S&P -0.17%, and the Russell -0.59%. The equal-weight S&P (RSP -0.23%) and the cap-weighted index (SPY -0.20%) moved almost in lockstep, showing this was not a narrow selloff dragged by a few mega-caps but a sector-level rotation.

2. The single strongest thing today was not any of the themes the pre-market brief bet on — it was precious metals: GDX +7.39%, GLD +4.14%, SLV +4.14%, IAU +4.12%. This was the largest single-sector move of the day, and the pre-market list did not cover it at all (the brief logged gold +2.28% only as a "dissonant signal," neither establishing it as a theme nor naming any beneficiary).

3. The precious-metals surge has a clear and verifiable explanation, and that explanation comes from today's two macro prints:

   • ADP July employment came in at just +44,000 vs. +70,000 expected (prior +98,000) — a large miss.
   • ISM Services PMI 54.1 (expected 54.5,6th month 54.0), but the internals were split: business activity 55.4 → 59.1, new orders 55.1 → 57.2 (both accelerating), while employment 51.2 → 47.4 fell into contraction and prices 67.7 → 70.3 (the fourth print above 70 in five months, with the 12-month average at its highest since April 2023).
   • "Employment collapsing + prices scorching" is a stagflation reading. It suppresses both rate-cut expectations (prices) and growth expectations (employment) — bad for equity valuations on both ends, good for gold on both ends. This also explains why the 10-year Treasury barely moved today (4.615%, -0.2bp) — weak employment should have sent bonds sharply higher, but prices at 70.3 pinned it down.

4. The most important judgment in the pre-market brief was validated at scale today: the brief's closing line, "beating expectations is no longer a sufficient condition in itself," was borne out by 5 mutually independent samples on the same day —

   • SNDK: revenue $8.97B (street $8.48B, a 5.8% beat), adjusted EPS $39.25 (street $34.96, a 12.3% beat), next-quarter guidance midpoint $10.55B also above the street's $10.47B → -5.40% on the day, then another -5.89% after hours;
   • WDC: EPS $3.56 (street $3.27, an 8.9% beat), revenue $3.75B (street $3.7B), next-quarter guidance EPS $4.00 vs. street $3.81, revenue $4.1B vs. street $4.04B (all above the street) → -11.20% after hours;
   • ALAB: yesterday's revenue +104%, Q3 guidance nearly +40% sequentially → -11.96% today;
   • AMD: revenue/EPS/guidance all three above expectations → -7.04%;
   • APP: revenue $1.92B, only a slight miss (street $1.94B), Q3 guidance 0.6% light → -16.30% after hours.

   What the five samples share: the size of the beat has decoupled from the direction of the stock. The market's current posture toward "high-expectation AI-linked assets" is: prove it to me, and even then I may not pay for it.

5. The pre-market list's directional hit rate was high: all 6 long recommendations closed green (LLY +4.86%, DIS +3.65%, WYNN +3.64%, ANET +3.57%, NVDA +3.43%, SHOP +16.98%); of the 14 "watch only / avoid" names, 11 did in fact decline. But the theme call was wrong — the brief ranked "AI infrastructure enablement layer" as its No. 1 theme, and today that theme fractured badly internally (ANET/NVDA up, ALAB -11.96%, MRVL -3.46%, SOXX -2.12%); it was not today's leading theme.

6. One major single-stock event the pre-market brief missed entirely: Alphabet (GOOGL) -4.03%, on Chief Scientist Jeff Dean's departure after 27 years (leaving alongside senior fellow Sanjay Ghemawat to found Discovery Loop), Demis Hassabis stepping back to DeepMind chairman, and Koray Kavukcuoglu taking over as SVP. Roughly $190 billion of market cap was erased in a single day, directly dragging down the communication services sector (XLC -1.04%).

7. Tone into tomorrow: VIX 15.81 (-4.18%) has fallen back below 16 — the market is not panicking; but tomorrow (8/6) brings the SPCX lockup expiry, a date-certain supply event, and the day after (8/7) brings July nonfarm payrolls. Today's ADP (+44,000) and ISM employment (47.4) are two consecutive prints pointing to downside risk for payrolls, while ISM prices at 70.3 makes the traditional reflex of "weak data = rate cuts = bullish" unreliable. Bias into tomorrow: defensives and precious metals continue; high-expectation AI/semiconductor assets stay under pressure.

---

1. Market Overview

1.1 Indices

The cause of the index divergence is clear: the Dow's constituents carry heavy pharma weight (AMGN +4.57%) and exclude GOOGL; the Nasdaq and S&P simultaneously bore the drag from GOOGL (-4.03%), semiconductors (SOXX -2.12%), AMD (-7.04%), and ALAB (-11.96%).

1.2 Market Breadth

This report's conclusion on breadth: the gap between equal-weight and cap-weighted was only 3bp, meaning "the average stock" and "the index" fell by the same amount today — today's pullback was not driven by a few heavyweights but was broad yet mild. This does not contradict the Dow's record high: the Dow is price-weighted and its sector composition skews defensive.

⚠️ This report did not obtain reliable advance/decline data, and has substituted the relative performance of equal-weight vs. cap-weighted ETFs as a breadth proxy. Raw advance/decline figures were inconsistent across sources in both methodology and date, so this report does not cite them.

1.3 Volatility, Rates, Dollar, Commodities

⚠️ Methodology note on gold's gain (important): CNBC's spot gold (XAU=) shows +0.07%, but that field uses a rolling 24-hour reference price, not the prior session's 16:00 ET close, and badly understates the day's move. This report uses gold ETFs settled at 16:00 ET as the basis for the day's gain: GLD 374.16 → 389.64, +4.14% (that closing price has been cross-verified as consistent across two independent sources, CNBC and Yahoo chart). Spot gold's intraday range ($4,157.18 open → $4,266.11 high) likewise confirms this was a large move up, not a flat day.

1.4 Sentiment Read

Today cannot be simply classified as risk-on or risk-off; the accurate description is "stratified":

• Not risk-off: VIX fell 4.18% to 15.81, the Dow set a record, XLF +0.21%, HYG (high-yield credit) -0.04% essentially flat — credit did not deteriorate.
• Not risk-on either: Nasdaq -0.83%, semiconductors -2.12%, precious metals surging (GDX +7.39%), GLD/SLV both +4.14%.
• The accurate description: capital exited high-expectation growth assets (AI/semis/high-multiple software) while flowing into defensives (health care XLV +1.27%) and real assets (precious metals, materials XLB +1.23%). This is not a reduction in risk appetite; it is a change in the type of risk — from "very long-duration growth stories" to "inflation-resistant real assets and defensives with certain cash flows."

This combination is fully self-consistent with the ISM data: employment 47.4 (growth has a problem) + prices 70.3 (inflation is unresolved) = buy gold, buy defensives, sell long-duration growth.

---

2. Pre-Market List Reconciliation

2A. Pre-Market "Long Recommendation" Bucket (priority deep-dive / watch closely)

Long-recommendation bucket, 6 names: 6 closed green, 0 closed red — 100% directional hit rate.

2B. Pre-Market "Watch Only / Avoid" Bucket

Watch only / avoid bucket, 14 names: 11 correct (fell, or should not have been bought), 3 errors (AMGN, GTE, NVO) — 78.6% hit rate.

2C. Hit-Rate Summary

2D. A One-Line Self-Critique (no hindsight bias — only what the brief actually got wrong)

An 85% directional hit rate masks a more important failure: the best-performing asset of the day did not get a single word in the pre-market brief.

Three specific points, ranked by severity:

1. Most serious: the brief saw the signal but failed to turn it into a conclusion. Section 8 of the brief, "risk signals," explicitly wrote: "risk assets and safe-haven assets rising together: gold +2.28%... this kind of divergence usually means some portion of capital does not believe the current risk appetite." The judgment was right, but it was filed under "risk disclosure" rather than "themes," and no beneficiary name was given at all. Result: gold rose another 4.14% that day and gold miners +7.39%, the strongest group of assets all day, with no entry point anywhere on the list. The brief demoted a "tradable observation" into a "disclaimer."

2. The theme ranking was wrong. The brief ranked "AI infrastructure enablement layer" No. 1, "GLP-1 stratification" No. 2, and "low-expectation repair" No. 3. The actual ranking today was: low-expectation repair (DIS +3.65%, WYNN +3.64%, both hit) > the long side of GLP-1 (LLY +4.86%) > AI infrastructure enablement layer (torn internally: ANET +3.57%, NVDA +3.43%, but ALAB -11.96%, MRVL -3.46%, SOXX -2.12%). The theme ranked third performed the most uniformly today; the theme ranked first did not even hold together as a "sector."

3. The way AMGN was downgraded was the problem (not the downgrade itself). The brief downgraded AMGN to "watch only" on insufficient evidence — downgrading when evidence is insufficient is correct discipline, and this report does not overturn that principle. But in the same list, AMGN's "total score 66" was still printed at No. 8 on the leaderboard, so the score and the conclusion contradicted each other: if the evidence is too thin to support a conclusion, it should not receive a score that invites cross-comparison. Today's lesson is not "we should have recommended AMGN," but "a name with insufficient evidence should not appear on the leaderboard carrying a score."

One point that needs to be stated clearly: the "avoid" call on VPG should not count as credit. It had already gapped -27.72% pre-market and closed -27.17%; the brief simply did not catch a falling knife — it did not predict anything. Counting it as a hit would inflate this report's scorecard.

---

3. Theme Validation for Today

3.1 Full Sector ETF Table (descending by change)

A detail worth recording: XLB (Materials) +1.23% ranked second, sharing the same driver as the precious-metals surge (the materials sector contains a large amount of metals & mining). The pre-market brief did not bring materials/mining into view at all.

---

4. After-Hours Earnings Moves (next-day catalysts)

Timestamp: POST_MKT quotes from 18:02–18:05 ET. After-hours liquidity is far lower than the regular session; the next open may deviate significantly.

4.1 The Most Important Item in This Section: the Pre-Market WDC Accounting Warning Was Confirmed by Official Numbers

The brief devoted an entire warning box to the argument that WDC's trailing PE of 30x was fake cheapness, because TTM net profit was badly flattered by fair-value remeasurement of its SanDisk stake (non-cash, non-operating), and on that basis cut its total score from 56 to 51.

Tonight's official numbers confirm this directly: for the same quarter, GAAP EPS was $8.21 while adjusted EPS was only $3.56 — a 2.3x gap. That is precisely the gap the brief pointed to.

What this means in practice for clients: any report tonight citing "WDC EPS $8.21, a huge beat" is comparing a GAAP figure against a non-GAAP consensus, and the conclusion will be exactly backwards. The street's $3.27 consensus corresponds to the adjusted $3.56, an 8.9% beat — not a 150% beat.

4.2 The Shared Structure of the Three "Beat but Fell" Samples

Placing SNDK, WDC, and APP side by side reveals a consistent and falsifiable pricing rule:

How to read it: beating expectations buys no upside, while a small miss costs 16%. This is a highly asymmetric payoff structure — under it, the expected value of holding a high-multiple name into earnings is negative, regardless of how good or bad the company's fundamentals are. The brief's "do not initiate positions ahead of earnings" on SNDK/WDC was therefore correct discipline, and the justification is stronger than what the brief itself wrote at the time.

---

5. Flows and Sentiment

5.1 The Direction of Sector Rotation

Inflows: health care (XLV +1.27%), materials (XLB +1.23%), precious metals (GDX +7.39%, GLD/SLV +4.14%), consumer discretionary (XLY +0.30%), financials (XLF +0.21%).

Outflows: energy (XLE -2.07%, XOP -3.93%), semiconductors (SOXX -2.12%, SMH -1.04%), communication services (XLC -1.04%), utilities (XLU -1.02%).

This was a rotation with clear logic, not indiscriminate de-grossing: what was sold was long-duration growth (semis, communications) and cyclicals hit by geopolitical de-escalation (energy); what was bought was inflation-resistant real assets (precious metals, materials) and defensives with certain cash flows (pharma).

5.2 Three Anomalous Signals That Must Be Read Together

1. VIX fell 4.18% to 15.81, yet gold rose 4.14%. These two usually move in opposite directions. Their occurring together means: the market is not worried about short-term volatility, but is buying insurance against medium-term purchasing-power risk. The brief was right to flag this divergence as a "risk signal"; today it widened further.

2. ADP came in at just +44,000 (vs. +70,000 expected), yet the 10-year Treasury did not budge (-0.2bp). Weak employment data should have sent bonds sharply higher. The reason it did not is ISM prices at 70.3 — the bond market believes inflation is unresolved and therefore dares not price in more cuts. Combined with the three officials who argued for a hike at the 7/29 FOMC, the old "bad news = good news" reflex is no longer reliable in the current environment.

3. Energy stocks fell far more than oil did. Brent +0.34%, WTI approx. -0.10%, while XLE -2.07% and XOP -3.93%. Equities are pricing an agreement that has not yet been signed (reports indicate the U.S., Iran, and Oman are close to a 60-day interim deal to reopen the strait toll-free). This implies that if the deal falls through, energy stocks have more rebound elasticity than crude itself.

5.3 Characterization

Today was not risk-off; it was risk-rotation. Credit did not deteriorate (HYG flat), volatility fell (VIX 15.81), and the Dow set a record — none of which supports a "flight to safety" narrative. But capital is systematically swapping duration out of the "AI growth story" and into "real assets + defensive cash flows", and what triggered this rotation was the "employment collapsing, prices scorching" combination in this morning's two macro prints.

---

6. Outlook for Tomorrow (2026-08-06, Thursday)

① Theme Continuity

② Tomorrow's Earnings and Macro Calendar

Macro (ET):

• 08:30 initial jobless claims: consensus roughly 195K–203K (sources differ), prior 197K. After today's ADP +44,000 and ISM employment 47.4, this is the last labor-market reading before this week's payrolls, and its weight is significantly elevated.
• 08:30 Q2 nonfarm productivity and unit labor costs: unit labor costs are a direct extension of today's "prices 70.3" narrative and are worth watching.
• ⚠️ The day after, 8/7 08:30, July nonfarm payrolls: consensus roughly +91,000, unemployment rate expected 4.2% → 4.3%. Both of today's employment readings (ADP +44,000, ISM services employment 47.4) point to downside risk.

Earnings (8/6): roughly 575 companies are scheduled to report. Larger-cap names confirmed by this report: COP (ConocoPhillips), SONY, DDOG (Datadog), TTWO (Take-Two), CEG, VST, KDP, KVUE, BDX, PH, HWM, CNQ, SRE, TRGP, QSR, RL, ATI, plus the previously flagged ABNB (implied volatility roughly ±8%).

⚠️ This report did not obtain the before/after-market timing split or consensus estimates for these names, and therefore gives no directional call on any of them. DDOG and TTWO, given their higher valuation multiples, are worth tracking first in today's "extremely low margin for error on high-multiple assets" environment — that is an inference from the environment, not a judgment on their results.

Date-certain events:

• 🔴 SPCX lockup expiry (8/6): roughly 911.5 million shares (approx. $116–123 billion) enter the sellable pool, with insiders able to sell at most 20% (roughly 182.3 million shares) in the first tranche. Musk's own roughly 6.4 billion shares are locked until 2027-06-12 and are not part of this tranche. SPCX is already -13.61% today on volume of 184 million shares (the largest in the market), stabilizing marginally at +0.87% after hours.

③ Key Focus for Tomorrow (ticker + validation point)

④ What to Avoid

1. High-multiple names reporting tomorrow (especially DDOG, TTWO) — five independent samples today already prove that in the current environment a beat buys no upside while a small miss costs double digits. The payoff structure is asymmetric.
2. Bottom-fishing the memory chain (MU / SNDK / WDC) — after SNDK delivered +372% revenue growth, an 84.6% gross margin, and a 12.3% EPS beat and still got hammered, "good results" is no longer a reason to buy this group. One must wait until the valuation compression stops on its own.
3. Trading the SPCX lockup expiry — after today's -13.61%, the risk is now two-way: it could rebound on "bad news exhausted," or genuine selling pressure could weigh on it. Neither the brief nor this report obtained insiders' actual intent to sell, and that uncertainty cannot be removed by analysis.
4. Chasing today's precious-metals gains — after GDX's +7.39% single-day move, a sound thesis does not mean a sound price, exactly isomorphic to today's SHOP lesson.
5. Directional bets on energy stocks (either direction) — the deal is unsigned and equities have priced ahead of the commodity; a single contrary statement would reset it.

⑤ Input Notes for Tomorrow's Pre-Market List

1. "Precious metals / inflation-resistant real assets" must be established as a formal theme, with names attached (GDX, GLD, SLV, plus metals & mining within the materials sector XLB). Today's lesson: the brief already saw this signal (the gold +2.28% divergence) but wrote it only into "risk disclosure" and not into "themes," so the strongest asset of the day had no entry point anywhere on the list.
2. Treat "pharma / defensives" as a sector rather than betting only on the GLP-1 industry line. AMGN has nothing to do with GLP-1 and still rose 4.57%, showing the buying is driven by defensive characteristics rather than the product cycle.
3. Tomorrow's brief must re-score the "AI infrastructure enablement layer" theme. A +3.57% to -11.96% split appeared inside that theme today; continuing to treat it as an S-grade sector theme will produce repeated errors, and it should be downgraded to "single-stock driven, no sector beta."
4. Establish a new screening rule and write it into the scoring: "holding a high-multiple name into earnings" should incur a direct score deduction. The payoff structure across today's 5 samples is highly consistent; this is not random noise.
5. Fix how names like NVO are handled: a name whose bad news has already been realized in the prior session should not continue to be tagged "avoid / short watch." One must distinguish "bad news not yet realized" from "bad news already priced."
6. At the execution level: names with insufficient evidence (like AMGN today) should not appear on the leaderboard carrying a cross-comparable total score. Either fill in the evidence, or list them as observations without a score.
7. The 8/7 payrolls print is this week's real watershed, and everything priced tomorrow should be treated as a pre-data interim. Today's ADP (+44,000) and ISM employment (47.4) are two consecutive readings pointing to payroll downside risk, but ISM prices at 70.3 means the traditional reflex chain of "weak data = rate cuts = bullish for stocks" can no longer be assumed to hold.

---

⚠️ Data-fetch failures and self-check log for this run (not for clients)

A. yfinance failed across the board, and the root cause differs from the one logged last time

• yf.Ticker('SPY').history() threw YFRateLimitError outright — zero successes.
• The root cause logged last time (a leftover local yfretry.py retry process hammering the IP) does not apply this time: ps aux confirmed no leftover data-fetch process other than this claude process.
• The workaround logged last time also failed this time: curl_cffi.requests.Session(impersonate='chrome') hitting query1.finance.yahoo.com still returned HTTP 429 Edge: Too Many Requests, indicating this is a sustained CDN-edge ban on this machine's IP, not a cookie/crumb-layer issue.
• ⚠️ Memory needs updating: the entry in yfinance-rate-limit-root-cause.md stating "the root cause is our own retry process" is not the only root cause; pure IP-level bans do occur, and curl_cffi impersonation does not work against them.

B. Working channels found this time (recommend hardening into the toolchain, with priority above yfinance)

1. quote.cnbc.com/quote-html-webservice/restQuote/... — the primary data source this run, zero failures throughout.
   • The multi-symbol separator is |, not a comma; passing a comma returns a silent error like {"symbol":"A,B","code":1} (no exception, no error message), which very easily produces empty data inside a scheduled job. Already hardened into work/fetch_cnbc.py / work/fetch_ah.py.
   • After-hours data lives inside the nested object ExtendedMktQuote (containing last/change_pct/volume/last_timedate), not in top-level flat fields. The first version of the script read top-level keys and got nothing back; fixed.
   • Index symbols: .SPX / .DJI / .IXIC / .RUT / .VIX; rates US10Y / US2Y; dollar .DXY; Brent @LCO.1.
2. query2.finance.yahoo.com is not banned (query1 is) — used this run to cross-verify GLD's 5-day closing series, returning 200. Next time yfinance is rate-limited, repointing to query2 should be the first option.

C. Data traps hit this run (all handled in the body)

• CNBC spot gold XAU='s previous_day_closing is a rolling 24-hour reference price, not the prior session's 16:00 ET close, which caused it to show +0.07% when the true move on the day was roughly +4%. This nearly produced a wrong "gold was flat" conclusion. Switched to the GLD/IAU basis (settled at 16:00 ET) and cross-verified with Yahoo query2's daily series (374.16 → 389.64).
  • The clue that triggered the investigation was that GLD/SLV/IAU all showed nearly identical gains (4.14/4.14/4.12), which looked suspicious at the time; verification showed those were real, and it was the "normal-looking" XAU= that was wrong. Lesson: cross-verification must target the key conclusions, not just the number that looks suspicious.
• @CL.1 (WTI) returned a stale 01:00 ET quote marked UNCH and is unusable as a closing price. The WTI close was taken from WebSearch results instead (approx. $75.69) and labeled "approx." in the body. @LCO.1 (Brent) has a normal timestamp and is usable.
• WDC's GAAP EPS $8.21 vs. adjusted $3.56 — a secondary source (gurufocus) put the GAAP $8.21 directly against the street consensus in its headline, when that consensus is actually a non-GAAP figure. Copying it verbatim would yield the absurd conclusion of a "150% beat." The body handles it on the adjusted basis ($3.56 vs. $3.27) and uses it to confirm the pre-market accounting warning.

D. Fetch failures

• WebFetch returned HTTP 403 for CNBC / TheStreet / invezz / tradingkey across the board (consistent with the previous report's log; now the norm). The related content was instead obtained via WebSearch summaries plus cross-referencing fetchable sites (prnewswire, fxleaders).
• prnewswire's official ISM press release fetches normally, and this run's six ISM sub-indices (especially employment 47.4 and prices 70.3) were obtained from it as primary data — these two numbers are the pivot of this report's entire macro narrative, which argues for adding prnewswire to the priority fetch list.

E. Data not obtained (left blank honestly in the body)

• Advance/decline: figures contradicted each other across multiple sources (one even gave "future" data for 8/6), so all were discarded; RSP vs. SPY was used as a breadth proxy with a note in the body.
• Specific earnings figures for Block (XYZ) and eBay: only after-hours quotes were obtained; marked "to be verified" in the body.
• The before/after-market timing split and consensus estimates for the 8/6 reporters: not obtained, so the body gives no directional call on them.
• Initial jobless claims consensus differs across sources (195K vs. 203K); the body gives a range rather than a single value.

F. Process log

• Per session instructions, the research sub-agents were not enabled this run (fundamentals-analyst / risk-auditor).
• ⚠️ Worth noting: the memory us-report-fundamentals-agents-worth-waiting.md records that the sub-agents once corrected a substantive valuation error. This piece is a recap (primarily reconciling against realized price action, with less fundamental inference than a pre-market brief), so the impact is relatively contained; but §4's SNDK/WDC earnings figures and methodology judgments are precisely where the sub-agents add the most value — this time it was caught by self-checking WDC's GAAP/non-GAAP gap, which means a human caught it, not that it will be caught next time. If sub-agents are re-enabled later, the recommended first use is methodology verification on after-hours earnings.

---

⚠️ Risk disclosure: this recap is an after-hours information review and observation only, and does not constitute investment advice. Data may differ in timeliness or methodology; please refer to company disclosures/SEC filings as authoritative. It must not be used directly as a basis for trading.