By Lily Caruso · Friday, September 4, 2026 — The Morning Brief, filed at 10:05 AM ET, fact-checked and revised at 10:20** --- Good morning. Pour the coffee. It's a four-day weekend for the market and a longer one for anyone who published a payrolls forecast this week, so let's be gentle with each other and precise with the numbers.

At 8:30 the Bureau of Labor Statistics reported that the economy added **162,000 jobs in August.** The consensus was 55,000 to 56,000. The entire range of forecasts Reuters collected ran from −25,000 to +121,000, and the print didn't beat the top of that range so much as walk past it by forty thousand. Unemployment held at **4.1%.** And then the line I'd put on the front page if I ran a newspaper: **July, first reported as a loss of 23,000 jobs, was revised to a gain of 21,000.** June was revised up as well.

Between them, the two prior months gained 55,000 jobs that hadn't existed on Thursday. Brian Jacobsen at Annex Wealth put it more plainly than I can: *"The drop in employment in July was an illusion. That's the problem with trying to trade the headlines.

Data get revised."* A month of this market — the soft-landing-turning-hard narrative, the rate-cut hopes, a good share of the small-cap selling — was built on a negative number that turned out never to have been negative. My colleague's desk had 60% odds on a print below 50,000, and printed them in public. That row is a loss, and it's logged as one in his column this morning in the same font as his wins.

I'll add only this. Every piece of evidence he cited was real: ADP said 38,000, JOLTS took its largest downward revision since 2025, and the services survey's employment index was in contraction. The evidence pointed one way and the number went the other.

The lesson isn't that the evidence was wrong. It's that August payrolls have two verdicts of their own, and we'll get the second one in October. **The first room: what's inside the number, because the composition argues with the headline.** The BLS wrote its own commentary this month, and it's better than anything on television. **Restaurants and bars added 59,000 jobs**, which the Bureau calls "well above the average monthly gain of 12,000 over the prior 12 months." Five times the run rate. **Local government education added 42,000**, "largely offsetting a decrease in the prior month" — the school-year swing. Those two lines alone are 101,000 of the 162,000.

Meanwhile **health care added 13,000**, which the Bureau itself describes as "a slower pace than the average monthly gain over the prior 12 months (+32,000)." That's the engine of this labor market running at forty percent of its usual speed. And **information lost 23,000**, "following losses that had averaged 8,000 per month" — nearly triple the run rate, in the sector that houses software, publishing, and the businesses most exposed to what AI does to headcount. The wage line is the one the hawks needed and didn't get. **Average hourly earnings rose 0.3% on the month and 3.1% over the year.** Peter Cardillo at Spartan Capital: *"There's no evidence here of rising wage inflation."* With both ISM price indexes above 70 this week, 3.1% is a real wage that isn't growing.

Participation rose to 61.6%. The broadest measure of underemployment, U-6, fell to 7.7% from 7.9%, and the number of people working part-time who want full-time work dropped by 414,000. Read it all together and you get a labor market that got *broader* in August, not tighter.

That distinction is the whole Fed argument, and it's why the smartest quote of the morning came from Jamie Cox at Harris Financial: *"This is a slider in the dirt for the Federal Reserve... There are several board members who will swing at this pitch to advocate for higher rates. The data support a hold, not a hike."* **The second room: the curve changed shape again, and this time the shape pays nobody.** Three days, three curves.

On Wednesday I told you the bank bid was a bear steepener — long rates rising faster than short — the one shape that pays a lender on the spread. On Thursday the mechanism flipped: yields fell, the bid got stronger, and I owed you the correction that banks make money on *volume* when rates stop rising. The deal calendar came back, Nvidia wrote a thirteen-billion-dollar check, and financials led.

This morning is the third chapter, and it's the unkind one. The **2-year yield rose 7.6 basis points to 4.41% while the 10-year rose 3.2 to 4.79%.** The front end led. That's a bear flattener — the shape the market draws when it's pricing a rate hike — and it's the one curve that helps neither the spread nor the calendar. **XLF was down 0.56% at 9:31**, from Thursday's leadership to the bottom half of the board in a single print.

And look at where the hit actually landed. At 9:35 the S&P 500 was down 0.05% and the Nasdaq was up 0.09%. **The Russell 2000 was down 1.40%.** Small caps are the room where the rate hike lives — floating-rate debt, domestic revenue, not much pricing power — and while the big index sat flat, the small one got sold. September hike odds moved to **59% from 52%** in the fed funds market.

The dollar rose 0.3% to 99.3. For the record at 9:42, my colleague's three rows for today: the 10-year at or above 4.78% is on track at 4.79%; hike odds above 55% are on track at 59%; and the S&P closing red is a coin flip, with SPY at $772.23, down 0.12%, and the Nasdaq doing the carrying. **The third room: gold failed the test it passed yesterday, and I'm glad it did.** Thursday morning gold rallied 1.81% on Waller's half-sentence while oil rose two percent on Hormuz, and I told you that was our wiring confirmed under laboratory conditions: war premium into the rate path, into real yields, into the metal. I also told you circuits run both ways.

This morning the test ran in reverse. A hot jobs print, hike odds up seven points, and **gold fell 1.7% to $4,392** on Reuters' stamp; **GLD was down 1.39% at 9:36.** Same war. Different Fed.

Opposite gold — in the other direction this time. Two consecutive days, two opposite inputs, two opposite outputs from the same circuit. A model that only works in one direction is a coincidence.

One that works in both is a mechanism. Bitcoin told the same story in miniature, up 3.2% to $81,224 before the print on Thursday's rate relief and down through $80,000 after it, per CoinDesk. The liquidity thermometer reads the same as the metal. **Four stories nobody is talking about, ranked by how much I think they'll matter by next Friday.** *One: the information sector just had its worst month of the year, and the software tape is already reading it.* Twenty-three thousand jobs lost in the one sector where "efficiency" is a product roadmap.

Now set that beside what the tape did to software this week. MongoDB fined. Palo Alto fined 9.5%.

Zscaler fined this morning after disclosing a 3% workforce reduction. Guidewire fined 15 to 19% for a guide 3% light. And Adobe, which named Anil Chakravarthy its next chief executive on Thursday afternoon, down 3% in the premarket six days before it reports.

Nobody is connecting the payroll line to the earnings line, and they're the same line. Software companies are cutting people and buying compute, and the market is fining the ones whose cash flow shows it. *Two: Zscaler confessed, in writing, that component inflation has reached software.* This is the sentence I'd underline in the whole week's filings. From Zscaler's Q4 shareholder letter: *"In Q4, we opportunistically accelerated certain purchases where data center equipment was available"* because of *"higher prices and tighter availability for memory, storage, and processors."* Memory, storage, processors.

That is Dell's constraint list — DRAM, NAND, CPUs — showing up two tiers down, in a cloud-security company's cash-flow statement, as the reason it bought hardware early before it got more expensive. It's why free cash flow was 7% of revenue in a quarter when operating cash flow was 31%, and it's the strongest confirmation the capex thesis I wrote last night could have received, from the company itself. The stock was paid 1.35% after hours, fined 5% in the premarket, and sat at **$172.30, down 3.09%, at 9:42**, off a low of $165.30.

The letter also answered my watch item: *"Security for AI"* bookings grew more than 50% quarter over quarter. So the product the capex is buying is selling. The market is grading the bill, not the product.

That's this week's statute, applied to a balance sheet. *Three: the marginal buyer of American duration is going home.* The yen is up 2.7% this week, its best since July. Japan's 10-year yield crossed 3% on Tuesday for the first time since 1996, and its 2-year is at a 31-year high. When a Japanese pension fund can earn three percent at home with no currency risk, it needs less of our 30-year, which sits at 5.25% this morning, within seven basis points of the 19-year high it touched on Tuesday.

Everyone is watching the Fed for the direction of long rates. The direction of long rates is increasingly set in Tokyo, and the Finance Minister's phrase this week — "heightened urgency" about the bond market — is the kind of language a country uses shortly before it does something. *Four: diesel just made an all-time high, and next Friday's CPI hasn't seen it yet.* $5.85 a gallon, per BNN Bloomberg. Brent is near $95 and on course for a third straight weekly gain.

The EU joined the U.S. sanctions campaign against Iran this week under a name — "Operation Economic Outcast" — that tells you how the people running it think about its duration. Diesel touches every delivered good, and it doesn't fully reach a consumer price index until a month or two after the truck pays for it. Which means the CPI that decides the September meeting may be the *friendliest* print of the autumn rather than the hottest, and the Fed knows it.

That's why Chair Warsh's language is about where inflation is running, not where it was. **The smaller rooms, quickly.** *Lululemon,* down 20% on a revenue decline and a second guidance cut, on a morning restaurants hired 59,000 people. The consumer is eating out and not buying leggings. That's a real-income story, and it agrees with the 3.1% wage line, not the headline. *Fair Isaac,* down 12% on what TheStreet describes as the FHFA ending the mortgage-scoring monopoly — this desk hasn't verified the specifics and says so — while Reuters has the 30-year mortgage at 6.71%, a one-year high.

Housing is paying the rate and the toll. *Volkswagen,* up as much as 9.7% in Frankfurt for cutting 50,000 more jobs, in the same week Zscaler was fined for cutting 3%. Layoffs get paid when they're the story and fined when they're a footnote to a guide-down. *Energy,* down 0.80% at the open despite the diesel record, which is the tape telling you it fears the Fed more than it fears the fuel. And *health care,* which comes with a correction.

Yesterday morning I told you the hiding place had emptied, with XLV down 0.47% at 9:54. By the close it had refilled: **XLV finished Thursday up 0.18%, at $173.26.** I graded a stamp, and I'm printing it. The Tuesday and Wednesday figures I cited were also morning stamps; the closes were +0.66% and +0.75%.

This morning it's down 0.47% again at 9:31, and the payroll table says the sector's own hiring engine is slowing — but the lesson of this week is that I don't get to call a room empty until four o'clock. **Next week, and why this morning dictates it.** Monday is Labor Day. Stocks and bonds are closed. So next week is four sessions, and the order matters more than usual, because this morning raised the stakes of every item on it. *Wednesday, September 9:* Apple's "Surprise and Shine" event at 10 AM Pacific. iPhone 18 Pro, a foldable iPhone that may be called Ultra — the first new form factor since 2017 — and an A20 Pro chip built on a 2-nanometer process, per MacRumors.

Read it as a TSMC story and a memory story before you read it as a phone story, and note that the foldable's price will test the same consumer who just fired Lululemon. The standard iPhone 18 is delayed to spring; the mix is all premium. *Thursday, September 10:* producer prices at 8:30 — the CPI preview — and then, after the close, **Oracle and Adobe**, witnesses five and six in the software courtroom, testifying after a week in which four of five were fined. Oracle is the AI-capex witness: the backlog, the cloud growth, the capital spending that funds it.

Adobe is the AI-displacement witness, now with a new chief executive attached. *Friday, September 11:* the August CPI at 8:30, which Governor Waller named as his decider and which now arrives with hike odds at 59%. Gary Schlossberg at Wells Fargo: *"Today's report going into the meeting certainly puts more pressure on the CPI to come in very friendly to avoid a rate increase."* Christopher Hodge at Natixis: *"The onus will continue to be on the doves to get a disinflationary print."* The meeting is September 15–16, and the blackout starts tomorrow. So here is the base case, and it's a window, not a day. **Next week is decided between Thursday at 4:05 PM and Friday at 8:30 AM** — Oracle and Adobe telling the market what AI is costing and displacing, then the CPI telling it whether the Fed can afford to care.

This morning moved the fulcrum. With hike odds at 59% instead of 50%, a friendly CPI is now the *surprise* and a hot one is the *confirmation*. If core comes in soft, Waller's half-sentence is restored in full: hike odds back into the forties, the 10-year back under 4.70%, small caps get their month back, gold retakes $4,500.

If it comes in hot, the 10-year retests Tuesday's 4.80% twenty-month high, the Russell breaks, and the market walks into the FOMC pricing the hike Warsh signaled at Jackson Hole. There's no middle case that matters. And Foxconn's August revenue usually lands around the fifth — this weekend, Taiwan time, if the pattern holds — which would make it the first read on whether the Dell and HPE order books are showing up in the factories before Tuesday's open. **The rows, in the same font as the loss above.** *One:* **60% that the Russell 2000 trails the S&P 500 from today's close through next Friday's close.** The mechanism is the one that showed up at 9:35: small caps carry the hike.

The falsifier is a friendly CPI, which un-prices it in a morning. *Two:* **55% that September hike odds on CME finish next Friday above 60%** — the market carrying today's 59% through PPI and CPI. The falsifier is the same print. I'm keeping these two deliberately correlated and saying so.

They're the same bet at two prices, and I'd rather be honest that next week has one variable than pretend it has two. *Grading note:* Episode 28's row — 60% that XLF beats XLU over five sessions from September 2 — grades at Wednesday's close, the fifth session, with Monday dark. *Watch, no row:* Apple's memory content per phone (the SanDisk line lives there), Oracle's backlog, Adobe's guide under a new CEO, and Thursday's PPI as the CPI preview. **Housekeeping, at full size.** Last night's security piece said Zscaler "was paid 4%." It was paid between 1.35% and 4% depending on the stamp, fined 5% by the premarket, and is unsettled until 4:00 — the headline caught the first verdict. The same piece carried Snowflake at 22%; the settlement was +16.55%. The XLV correction is above, in the room where it belongs.

The first draft of this episode said Adobe's CEO news was announced "today"; it was Thursday at 1:05 PM. My colleague's column this morning corrects its own first draft on the CPI date — Friday the 11th, not Thursday — and re-grades the HPE row from loss to win, because we graded a stamp instead of the $54.44 close. The decoy chronicle stands at 60: a June earnings calendar surfaced as next week's, and two old Foxconn revenue reports, from last September and this March, were served as August's.

Thrown out, and printed so you know. One image before the weekend. In early August the Bureau of Labor Statistics published a number that said the economy had stopped hiring.

Markets traded it for a month. This morning the same agency said: never mind, it was hiring after all — and then added 162,000 on top. Even the data has a first verdict and a second one.

The market is doing the only reasonable thing you can do with that knowledge. It's trading the second verdict on July, and waiting for the second verdict on the Fed. That one arrives next Friday at 8:30, in a report that hasn't seen the price of diesel yet.

See you at midday. Enjoy the long weekend — the bond market is. — Lily **Tickers in play:** IWM · SPY · QQQ · XLF · XLV · XLE · GLD · TLT · ZS · ADBE · ORCL · AAPL · TSM · SNDK · MU · LULU · FICO · GWRE · HPE · DELL · FXY --- *This is TrendyVest's analysis and opinion — for informational purposes only, not investment advice or a recommendation to buy or sell any security or commodity. Sources: the August 2026 Employment Situation (payrolls +162,000; unemployment 4.1%; June revised +20,000→+31,000 and July −23,000→+21,000, net +55,000; average hourly earnings +0.3% and +3.1% over the year; participation 61.6%; food services +59,000 "well above the average monthly gain of 12,000 over the prior 12 months"; local government education +42,000 "largely offsetting a decrease in the prior month"; health care +13,000, "a slower pace than the average monthly gain over the prior 12 months (+32,000)"; information −23,000 "following losses that had averaged 8,000 per month"; part-time for economic reasons −414,000) per the Bureau of Labor Statistics release, Sept 4, and the October 2 date of the next report per the BLS calendar; U-6 7.7% from 7.9% per investinglive, Sept 4; the 56,000 consensus and the −25,000 to +121,000 range per Reuters (Mutikani), Sept 4, via Yahoo Finance; the 2-year (+7.6 bp to 4.41%), 10-year (+3.2 bp to 4.792%), 30-year (5.252%), dollar index (+0.3% to 99.3), gold (−1.7% to $4,392), the 59%-from-52% hike probability, and the Cardillo and Jacobsen quotes per Reuters' Instant View via Investing.com, Sept 4; the Cox, Schlossberg and Hodge quotes per Reuters via Yahoo Finance ("Strong August jobs report sends yields higher"), Sept 4; the 9:35 AM index stamps (S&P −0.05%, Dow −0.16%, Nasdaq +0.09%, Russell 2000 −1.40%), Lululemon −20.4%, Guidewire −14.5%, Fair Isaac −12.2% and the FHFA characterization, Volkswagen and the 50,000 job cuts, and the Iran sanctions item per TheStreet's Sept 4 live blog; ETF stamps per stockanalysis.com, Sept 4 (SPY $772.23 −0.12% at 9:42; QQQ $719.77 +0.29% at 9:31; XLF $58.23 −0.56% at 9:31; XLV $172.44 −0.47% at 9:31; XLE $64.10 −0.80% at 9:31; XLU $43.00 −0.07% at 9:32; XLK $186.82 +0.46% at 9:31; GLD $404.50 −1.39% at 9:36), XLV's closes (Sept 1 $171.67 +0.66%; Sept 2 $172.95 +0.75%; Sept 3 $173.26 +0.18%) per stockanalysis.com's history page, and Zscaler ($172.30 −3.09% at 9:42, range $165.30–172.66; after-hours $180.20 +1.35% at 4:49 PM Sept 3; premarket $169.00 −4.95% at 8:38) per stockanalysis.com; Bitcoin ($81,224.60 +3.23% pre-report per TheStreet; below $80,000 after the report per CoinDesk, Sept 4); Zscaler's shareholder-letter quotes ("opportunistically accelerated certain purchases," "higher prices and tighter availability for memory, storage, and processors," "approximately 3% of employees," and "Security for AI" bookings "50%+" Q/Q) per Zscaler's Q4 Fiscal 2026 Shareholder Letter, Sept 3, and OCF $279.3M (31%) / FCF $60.8M (7%) per the Sept 3 release; Adobe's leadership announcement (Anil Chakravarthy to become president and CEO December 1, 2026; Shantanu Narayen to executive chair; announced Sept 3, 1:05 PM) per Adobe's newsroom, and Adobe −3.18% premarket per Investing.com/Reuters, Sept 4, 5:06 AM; the yen +2.7% weekly (best since July) per Yahoo Finance's Sept 4 live blog, the Katayama "heightened urgency" line per FXStreet, Sept 4, and Japan's 10-year at 3% (first since 1996), 2-year 1.81% (31-year high), the U.S. 30-year 5.272% close and 5.323% 19-year high, and the 10-year's 4.798% 20-month high, all Sept 1, per Traders Agency's Sept 1 report citing CNBC, Yahoo Finance and Reuters; diesel $5.85 all-time high per BNN Bloomberg, Sept 4; Brent's third weekly gain per FinancialJuice, Sept 4; the 6.71% mortgage rate per Reuters, Sept 4; the Labor Day closure of stock and bond markets per Yahoo Finance, Sept 3; Apple's Sept 9 event (10 AM PT; iPhone 18 Pro/Pro Max; foldable; A20 Pro on 2nm; standard iPhone 18 to spring 2027; first new form factor since the 2017 iPhone X) per MacRumors, Aug 26; Oracle's Q1 FY27 release Thursday Sept 10 after the close per the company via StockTitan; Adobe's Q3 FY26 release Sept 10 per Business Wire, Aug 31; PPI Sept 10 and CPI Sept 11 per the BLS release calendar; the Sept 15–16 FOMC and Waller's decider per Reuters and this desk's Episode 29; Lululemon's Q2 (revenue ~$2.4 billion, −4%; guidance lowered again) per BNN Bloomberg, Sept 3; Guidewire's Q4 and Q1 guide per Reuters via Yahoo Finance, Sept 3; Foxconn's monthly-revenue timing per the company's prior monthly releases; Wednesday's and Thursday's XLF, XLV and curve readings per Episodes 28 and 29; the payrolls row, the HPE re-grade and the CPI-date correction per this desk's Sept 3–4 Premarket Edges and the master tracker.

Stamps were moving at publication; settlements are sworn at the close. Do your own research.* *Markets. Tech.

The Edge. Research with receipts.*