The third quarter opens with the tape holding a question it did not answer before the long weekend: the last two sessions of June rotated hard out of the crowded artificial-intelligence names and into gold, silver, and the short end. The only question that matters this morning is whether the market comes back Monday wanting to finish that rotation — or to reverse it.

Where the Tape Was Left — Friday, July 2, 2026 Settled Close

This is the last close of record. United States markets were closed Friday, July 3, for the Independence Day holiday (observed). The figures below are not July 6 marks; no July 6 prices exist yet at the time of filing.

Sources: Yahoo/yfinance batch, two-source confirmed to the cent against the Yahoo v8 chart Application Programming Interface (“API”); United States Treasury (“UST”) par yields from the Treasury.gov daily par curve. Marks are the Friday July 2 settled close.

Section I — The Question the Long Weekend Left Open

The third quarter and the new trading week open on an unresolved tape. The final two sessions of June carried a clear and deliberate message: on Wednesday, July 1, and Thursday, July 2, the market rotated out of the crowded artificial-intelligence complex and into stores of value and the short end. The signature was unmistakable — on July 2 the Dow Jones Industrial Average printed a record 52,900.07, up 1.14%, while the Nasdaq Composite fell 0.80% to 25,832.67, and gold and silver ran to fresh highs of $4,193.70 and $62.76 even as the VIX fell to 16.15. A record in the old-economy index against a decline in the growth index, on a low-volatility day, is the textbook signature of an orderly rotation rather than a fear event. The market did not sell risk; it rotated the kind of risk it wanted to hold.

The catalyst was the June employment report, released early on July 2 ahead of the holiday. Nonfarm payrolls rose just 57,000 against a consensus near 113,000, with April and May revised down a combined 74,000, and the unemployment rate ticked down to 4.2% for the wrong reason: the labor-force participation rate fell to 61.5%, its lowest since March 2021, as roughly half a million people left the labor force. A soft headline with a shrinking labor force is precisely the print that trims the odds of further Federal Reserve tightening — supporting duration and gold — without handing the growth names a clean growth story. That is the macro backdrop the market carried into the three-day weekend, and it is the backdrop it must now either extend or reject.

So the question this morning is narrow and falsifiable. Either the rotation resumes — the crowded artificial-intelligence names continue to give back the extraordinary gains of the second quarter while the metals and the short end keep their bid — or it reverses, with dip-buyers treating the two-day pullback as a discount on the only trade that has worked all year. The publication does not pretend to know which. What it can do is lay out the evidence on each side, the calendar that will settle the question, and the way the Operation Epic Fury (“OEF”) book is positioned while the market decides.

Section II — The Case for Resume, and the Case for Reverse

The case that the de-risking resumes. The bull market in the artificial-intelligence complex spent the second quarter reaching valuations that price permanent scarcity and infinite demand. Corning ran 391% over the trailing twelve months into the June 30 close; Micron crossed a trillion dollars in market capitalization; the Magnificent Seven and their memory-and-accelerator supply chain carried the entire index. The soft June jobs print is the first crack in the growth story that justified those multiples, and the rotation into gold and the short end says the smart money is repositioning for a second half in which the labor market rolls over and the Federal Reserve’s next move is more likely a cut than a hike. In that world, the duration-and-quality trade is the trade, and the crowded growth names keep giving back. The metals bid — gold and silver at fresh highs on a low-VIX day — is the tell that this is insurance-buying, not panic, and insurance-buying tends to persist.

The case that the de-risking reverses. Against that, the reversal case is simple and has been right all year: every pullback in the artificial-intelligence trade since the cycle began has been a buying opportunity, and the second-quarter earnings season that starts this week is far more likely to confirm the demand than to break it. FactSet forecasts S&P 500 earnings growth of roughly 22% for the second quarter — the second straight quarter above 20% — and the positive revision pattern into the season is itself unusual. If the banks on July 14 confirm healthy credit and capital-markets activity, and if the first mega-cap technology guidance in late July reaffirms the capital-expenditure build-out, the two-day de-risking will look like ordinary pre-holiday profit-taking in thin volume, and the dip-buyers will be paid. A record Dow is not a risk-off signal; it is a market that wants to stay invested. The reversal case is that it simply rotates back into the leaders once the earnings remove the doubt.

The publication’s read. The honest synthesis is that this is a genuinely two-sided setup, and the tell will be behavioral, not fundamental, in the first few sessions. Watch whether the metals hold their new highs on the return to trading: if gold and silver give back the July 2 spike as equities reopen firm, the reversal is winning and the rotation was a long-weekend artifact. If the metals extend while the Nasdaq stalls, the de-risking is real and the second half has changed character. The book is deliberately built to be paid either way at the margin — long the constraint themes that do not depend on the artificial-intelligence multiple, short the most-overbought semiconductor name through the clean P39 expression, and already having banked the most extended long positions at the quarter-end high. That last decision is worth quantifying precisely, which is the subject of the next section.

Section III — The Harvest Shadow: What the June 30 Closures Did and Did Not Give Up

A recurring discipline begins with this edition. When the book harvests a position, the decision is only half-testable at the moment of sale: the realized profit is banked, but whether the sale was well-timed depends on what the position would have done had it been held. The publication will now track that counterfactual — the “harvest shadow” — for the ten pairs closed at the June 30 settled close, marking each still-held position forward and comparing it to the profit actually booked. The first reading is taken at the July 2 settled close, and the exercise will be repeated through July as the market wanders.

The method is straightforward. For each closed pair, the realized profit-and-loss booked on June 30 is compared to what the same position — identical share counts, identical entry basis — would be worth marked to the July 2 close if it had never been sold. A negative “shadow delta” means the position was worth less on July 2 than the profit already banked: the harvest was well-timed and the sale avoided giving back gains. A positive delta means the position would have been worth more if held: profit was left on the table.

All still-held marks are the July 2 settled close, two-source confirmed to the cent (yfinance batch plus Yahoo v8 chart API). Exact inception share counts throughout. Green in the delta column denotes a sale that avoided a give-back (booked better); red denotes profit left on the table.

What the first reading shows. The June 30 harvest banked $595,592 across the ten closures. Marked to the July 2 close, those same positions would be worth $434,507 — a difference of −$161,084. In plain terms, booking the trades at the quarter-end high was roughly $161,000 better than holding them into the first two sessions of July. The eight profit closures, all in the crowded artificial-intelligence names that led the sell-off, were each worth materially less by July 2: Corning-versus-Microsoft alone (P1) gave back roughly $76,000 of paper value in two sessions after it was sold, and every one of the eight would have shed gains if held. That is the harvest working exactly as designed — the concentration names were sold at their peak, two sessions before they rolled over.

The honest asterisk. The two exceptions are the SOXS liquidations. The leveraged-inverse semiconductor pairs (P35 and P37) would together have recovered roughly $97,000 if held, because SOXS bounced from $3.24 to $4.51 while the short legs fell. Read narrowly, that is profit left on the table. But those legs were closed for a structural reason, not a directional one: the leveraged-inverse decay that corrupts a SOXS position’s profit-and-loss made them unsuitable to hold regardless of the two-day bounce, and the same thesis was re-expressed cleanly in the new P39 (long Broadcom, short Intel) without the decay. The harvest shadow is a scorekeeping tool, not a second-guessing one: it records that the SOXS pairs bounced, while the reason for closing them — removing a structurally broken instrument from the book — stands independent of that bounce. Net of both effects, the quarter-end sweep was −$161,084 better than inaction, and the discipline of tracking the shadow will show, over July, whether that verdict holds as the market wanders.

Section IV — The Road to Labor Day: A Calendar for the Resume-or-Reverse Question

The question posed at the top of this edition will not be settled in a session. It will be settled by two streams of information arriving between now and Labor Day: the macro data that tells the market whether the soft June labor print was a blip or a trend, and the second-quarter earnings that tell it whether the artificial-intelligence build-out that drove the whole rally is still compounding. The calendar below maps both streams through Monday, September 7. Where a date bears directly on a book position, it is flagged. This is the section that turns “resume or reverse” from a mood into a schedule.

Macro release dates: Bureau of Labor Statistics and Federal Reserve published schedules (CPI June on July 14, July on August 12, August on September 11; the Employment Situation for July on August 7; the September FOMC on September 15–16 carries the next Summary of Economic Projections). Earnings dates from company guidance and season previews; confirm individual names on a live calendar as the season approaches. The three tells to watch for the spine: the money-center banks (July 14), the first hyperscaler capital-expenditure guidance (Microsoft and Meta, July 29), and Nvidia (late August).

Section V — Themes Callback: the Constraints Do Not Care Which Way the Trade Breaks

The resume-or-reverse question is about the artificial-intelligence multiple — whether the market keeps paying up for the growth names. The constraint themes the publication laid out in the July 1 quarter review are about the artificial-intelligence physical build-out, and the distinction matters this morning because the constraints bind regardless of which way the crowded trade breaks. A book positioned in the constraints rather than the multiple is insulated, at the margin, from the very question that will whipsaw the momentum names.

The aviation-fuel warning still governs. The jet crack spread — the refining margin over crude — held north of $40 to $50 per barrel into the July 2 close even as crude sat below $70, more than double its historical norm near $20. That gap is the tape saying the physical refining and logistics damage from Operation Epic Fury has not been repaired, whatever the crude headline says. The book expresses it through the airline shorts (JetBlue in P24, Delta in P31, International Consolidated Airlines Group in P33) and the product-tanker long (Scorpio Tankers, the long leg of P33). Delta’s report on Friday, July 10, is the first live test: unhedged fuel cost is the line to watch, not the revenue headline.

The memory demand-sink is where resume-or-reverse actually resolves. Every constraint — helium for lithography cooling, grain-oriented electrical steel and gas turbines for power, copper for wiring — ultimately feeds the memory-and-accelerator complex. That complex is priced for the constraints to bind permanently, and it is exactly the crowded trade the July 2 rotation sold. The mirror image of the aviation warning applies: just as the crude tape can lie about the war being over, the memory tape can lie about the build-out being infinite. The book’s answer was to harvest the most extended memory-adjacent longs (Corning, Micron, Generac) at the June 30 high — the harvest shadow above confirms that was well-timed — while keeping the clean relative-value expression through P39 (long Broadcom, short Intel). Microsoft and Meta on July 29 and Nvidia in late August are where the demand-sink either confirms or breaks.

The sovereign thread is the tail nobody is pricing. The metals bid that led the July 2 rotation is not only a growth-scare trade; it is the same monetary-credibility hedge the publication has tracked through the Global Sovereign Credibility Event and the gilt-to-Treasury transmission mechanism. If the soft labor data pulls forward expectations of Federal Reserve easing while the long end refuses to rally — because the fiscal arithmetic will not allow it — the gold and silver highs of July 2 are an early print of that tension, not a passing risk-off. This is the thread that would make the de-risking structural rather than tactical, and it is the one least visible in the daily tape.

Section VI — Pair Book at the Last Close of Record (24 active pairs, July 2 settled)

This is the book of record carried into the July 6 open, marked to the Friday July 2 settled close — not to any July 6 price. Book state is the v5 canonical post-June-30-sweep: 24 active pairs, 15 closed. Ranked by pair P&L versus entry. P27 (PKX/MT) closed June 12; P38 (PKX/SLX) active; P23 (DAL/CRWV) closed May 29. P5 short leg carries MMT (Aberdeen Multi-Market Income Fund, converted from MCR effective June 22 at ratio 1.33352869; basis $4.5443). The first fresh marks of the third quarter will be taken at tonight’s July 6 settled close and carried in the next edition.

Book totals (July 2, 2026 settled close — carried into the July 6 open):

• Unrealized P&L (24 active pairs): −$191,955 (6 pairs positive, 18 negative)
• Realized register (15 closed pairs): +$487,933.25
• Inception-to-date (“ITD”): $295,979

Exact inception share counts throughout; notional approximation prohibited. All marks are the Friday July 2 settled close, two-source confirmed to the cent. The book was not marked over the July 3 holiday and is not marked to any July 6 price in this premarket edition.