1The unemployment rate fell because the workforce shrank — and the hike got pushed out, not priced out
The July Employment Situation printed at 08:30 Eastern and repriced the whole board. Nonfarm payrolls contracted by 23,000 on the month, after adding 20,000 in June. The unemployment rate fell anyway, to 4.1% from 4.2%, and it is now below the 4.3% of May. Those two facts fit together only one way: participation dropped to 61.4%, four-tenths of a point below its April reading, so the labour force shrank faster than the jobs did.
The rest of the release points the same direction. Average hourly earnings rose to $37.62 from $37.60 — 0.05% on the month, roughly 0.6% annualised, which is wage pressure that has essentially stopped. Underemployment was unchanged at 7.9%, so nothing is deteriorating out of sight beneath the headline. The Sahm rule reading went negative, to −0.03 from +0.07, though that is mechanical rather than reassuring: a falling unemployment rate suppresses it regardless of what payrolls do.
The pricing followed within hours. Polymarket's September Fed contract — the deepest on the board at $18.5m of volume and $2.19m of liquidity — moved to 62.5% for no change, up fourteen points on the day, against 36.5% for a 25bp increase, down eleven. A day earlier the same contract was a coin flip at 48.5 against 47.5. A cut is barely in it at about 2.8%, and did not move.
The year is a different market. Polymarket's contract on a rate hike landing at some point in 2026, at $6.87m of volume, still prices one at 54.5% — down nine points on the day, and still better than even after the softest payrolls print of the cycle. Read the two together and the hike is being pushed out rather than priced out.
The mechanism is what makes that coherent. A September move was aimed at wage-push inflation, and wages are rising at 0.05% a month, so the channel it was meant to lean against has stopped transmitting. The hawkish counter — that a shrinking labour force is supply destruction, and therefore inflationary at the margin — survives the print intact. It just acts slowly, and nothing before next week's CPI can test it.
2Reported profit at the biggest AI spenders has stopped measuring the business
Aswath Damodaran, on Prof G Markets, makes a narrower argument than his episode title suggests. His claim is not that the spending is unaffordable — he thinks they can afford it. It is that the Mag 7 have quietly become capital-intensive manufacturers, so return on invested capital against cost of capital is now the metric they live or die on, and not one of them has ever been judged by it. Five years ago, at returns of 70–90%, invested capital was irrelevant.
The accounting is where it bites. Eighty-five per cent of Amazon's net income and 87% of Alphabet's came from unrealised markups on private stakes — Anthropic, OpenAI, SpaceX — rather than from operations. A price-to-earnings ratio below 20 therefore looks cheap without being cheap, because the numerator and the denominator have stopped describing the same thing. On the revenue side he cites third-party estimates that OpenAI and Anthropic together account for 73% of 's AI revenue, per Barclays, and 74% of Microsoft's, per Wells Fargo — architecture revenue rather than end-customer revenue.
He has acted on it, selling and . The Nvidia objection is specific: roughly $500bn of guarantees for OpenAI chip purchases and for client land and buildings, which means valuing the company now requires valuing OpenAI too. Hold the caveat in the same breath as the call — he still owns five of the Mag 7 and describes himself as an optimist on AI as a product, so this is a complaint about unreadable accounts, not a bubble call. He expects no single break either: staggered pain closer to 2001 than to one crash, with capex write-offs and restructuring charges as the first real admission. gets a favourable mention as one of the few names able to show money made selling AI-embedded product to end customers, though he states no position.
The disclosure gap is corroborated from a different seat. Alex Pollak of Loftus Peak, on Equity Mates, puts roughly $1.6tn of off-balance-sheet obligations across five large technology companies, falling due in three to five years for GPUs, buildings and power — disclosed in the notes rather than on the balance sheet, and up significantly in twelve months. carries about $420bn off-book against $140bn on it. is the outlier that shows the size of the bet its peers have made: $11bn of capex against $150–200bn at those peers, a deliberate wait-and-see that still pays a side cost in higher chip prices and has already forced device prices up.
So the number that decides these companies has changed, the disclosure that would let a reader check it sits in the notes, and on Damodaran's own account the first hard evidence arrives as a write-off rather than as a guidance change.
3The bond market and the equity market are pricing the same AI build-out differently
is where the two prices sit furthest apart. Pollak's evidence is that recently issued Oracle bonds trade about 10% below issue price four weeks after issue, and that its credit default swaps are at their highest since the financial crisis. The mechanism is refinancing arithmetic: if the spread goes from 100–150bp over treasuries to 200–300bp, the economics start to look like junk and the build stops paying for itself. That is the third warning on the same name from a third channel in three days.
The striking part is who agrees. Ed Yardeni, on The Compound, is the most bullish of the day's four videos by a distance — the S&P 500 at 8,250 by year-end and 10,000 by the end of 2029, built on $413 of 2027 consensus earnings at a 20x multiple, with no recession before 2030 — and he concedes the credit point without argument. Hyperscaler spreads over treasuries are widening, Oracle's swaps are at a multi-year high, and the debt market, he says, sees problems before the equity market does.
His equity case is testable, which is what makes it worth setting against that. Multiple expansion over three years has been 18%, against 250% into the dot-com peak, and this quarter's earnings growth is 28.8% with 15% margins — earnings-led, not multiple-led. The signature he points to is genuinely odd: the six-month forward P/E is down 8% while the index is up 11%, where an 8–10% multiple contraction would normally arrive with a 9% price fall. Breadth supports him — a record high in the advance-decline line, the other 493 names beating the Mag 7 this year, small caps participating after sitting out two years of the bull market, and banks bought into this week's melt-up (, ). The tape agreed on the day, with the Russell 2000 up 1.00% against 0.45% for the S&P 500 (). He also cut technology and communication services to market weight last December and has not moved since, so his bull case is not a mega-cap case. Take the bias with it: he describes himself as permanently bullish and says he reads bears specifically to find what he is missing, which is a stated direction of error rather than a hidden one.
The equity market is not uniformly generous to the build-out either. reported $3.2bn of quarterly profit against $282m a year earlier and fell 15%, which the wires put down to AI capex worries rather than to anything in the print. and are named as the pricing-power engine behind that 28.8% earnings figure — selling out of supply and raising unit prices — with the caution attached in the same breath: high prices invite innovation, so they may not be obvious holds by year three or four.
Oracle's next refinancing is where this gets settled, because it is the one place the two markets have to agree on a number.
4Gold's breakout is being driven by buyers who do not care about the price
The setup Ross Givens describes on his own channel, in a video titled "The Gold Explosion Has Started... And It's NOT Too Late", is technical — Bollinger band width on gold compressed to its tightest reading since August 2025, then a single-session break from about $4,100 to about $4,300 on above-average volume, the best day since February and the best week since January. The demand he describes behind it is not technical at all.
Central banks bought 289 tonnes in the second quarter, the largest Q2 the World Gold Council has recorded. Poland has taken 82 tonnes year to date, to a record 632, against a stated 700-tonne goal; Uzbekistan added 41. Chinese gold ETFs took inflows for fourteen consecutive sessions after two months of outflows, one day above $2bn, on domestic conditions rather than a global trade — falling property prices and capital controls leave savings few places to go. China also banned paper gold trading on 24 July, so the Shanghai Gold Exchange now settles in physical metal, and Hong Kong airport vault capacity is expanding from 150 tonnes to 1,000.
Two things travel with that. Givens corrects himself on camera: the World Gold Council has reclassified part of the 244-tonne first-quarter central-bank figure he previously cited into over-the-counter institutional buying. That is a credibility marker and it is rare. But the video is also a funnel — a paid membership is pitched twice inside thirteen minutes — and the price targets ride with the pitch: above $6,000 in 2027, $10,000–12,000 over two to three years. The trade he actually describes is far more modest and far more disciplined: entry near $4,300 spot, a stop below the summer swing low around $4,060, and a hard exit under $4,000. His own position is a micro gold futures contract, with and named as the retirement-account expression of the same view — same custodian, differing on price and expense ratio. Yardeni, separately, has raised his year-end gold target to $5,500, while conceding that gold lends itself to technical rather than fundamental analysis.
Reserve managers and savers escaping capital controls buy on a schedule rather than on a chart. That is a different marginal buyer from the one a breakout usually attracts, and a slower one to leave.
5The Strait of Hormuz is being marked down while the war is being declared over
Polymarket's contract on Hormuz traffic returning to normal by 31 December fell four points to 58.5% on $7.33m of volume, down from 62.5% a day earlier. The wire supports the direction: Reuters reports transit dwindling this week, a proposed passage arrangement that shipping sources judge unfeasible, an Iranian parliamentary committee reviewing a bill to bar US and Israeli ships, and Saudi officials expecting an imminent two-pronged attack — days after a Saudi–Turkey–Pakistan mutual-defence pact was signed.
The White House is putting the opposite case. The same article, headlined "Donald Trump Won The Iran War", was posted or re-posted on Truth Social three times inside fifteen hours. The repetition is the signal, and it points against the market.
Crude gained on the session regardless, despite a payrolls print that was soft on growth. On a day when growth expectations fell, that only reads as a supply-risk bid.
What the sources recommended
Nine names carried a direction across the day's four videos, and six of them come from one fund manager describing his own book on an episode his fund sponsored.
The buys. and are Ross Givens' retail expression of the gold trade above; he is running it himself through micro futures, which is not the same instrument. is Alex Pollak's first-choice biotech holding, on the view that the Colonia acquisition buys a platform for treating cancer in the body rather than a single mechanism — part of a deliberate push of biotech past a tenth of the fund. and come from the same rule that produces most of his sells: he will not underwrite revenue beyond eighteen months to three years, and both clear it, Alphabet on Google Cloud revenue up 75% with a backlog over half a trillion dollars, Microsoft on its net cash position. One caveat covers all four — the episode was sponsored by his own fund, disclosed at the top and in the outro, so these are disclosed positions rather than independent analysis.
The sells. and are positions Damodaran has exited, the Nvidia case resting on the guarantees described above. was named twice in the same window, because Pollak has avoided it for roughly eight years on the visibility rule — the valuation needs a ten-year revenue forecast he will not make. is the bond evidence. is a decline-to-buy on that same rule; Damodaran separately calls the IPO too sloppy and reckons about 80% of the value is really an xAI story embedded in a launch company, without stating a position.
Two things that look like calls and are not. Nvidia, , , and are all still in Pollak's book, held through a semiconductor drawdown of more than 20% rather than added to — which puts the same Nvidia on both sides of one day's calls, sold by one manager and sat on by the other. And the memory names are a caution inside a bullish frame, not a call either way.
Wednesday's CPI can undo the whole day
The September repricing rests on a single reading: that wage-push inflation has stopped transmitting. July CPI lands on Wednesday 12 August and is the first thing that can test it. Polymarket's contract on the annual rate barely moved on the jobs data, which is the correct behaviour — payrolls do not forecast next week's prices — and it sits at 3.4% modal on 44.5%, with 3.3% at 31.5% and 3.5% at 17%. Treat that lightly, because the market is thin: $383k of total volume and about $2,500 a leg over 24 hours. The bar it sets is clear enough anyway. A 3.5% handle puts the September hike straight back on the table and unwinds the fourteen-point move.
Two of the day's four videos were recorded before the payrolls print — the Equity Mates episode published yesterday evening, The Compound's this morning — so their rate commentary is one print behind. Only the gold video is post-print, and that trade carries its own dated test: a break of $4,000 ends it on its author's stated terms, no interpretation required.
