1A September hike is a coin flip, and the case for it rests on energy rather than wages
Polymarket's September Fed book is the deepest contract on this page, and it has stopped choosing. A 25bp increase prices at 49.5% and no change at 49.5%, on $592k and $582k of liquidity and $206k and $304k of 24-hour volume. A day earlier those legs read 50.5% and 48.5%. The hike gave a point back over Saturday, having gained three across the week. The two outcomes are now exactly tied.
Joseph Wang, on Fed Guy, walked the week's chronology and it explains the oscillation. The chair's Jackson Hole speech, he said, "all but promised a September hike", and pricing sat near 60%. Governor Waller's remarks then read dovish and took it to 50/50. The August payroll print pushed it back to just under 60%. Waller's condition was explicit: he "explicitly told everyone that if CPI came in hot next week, he would be open to hiking in September", while arguing that tariff-driven inflation is fading and asking, in his own paraphrase of John Lennon, to "give disinflation a chance."
What Wang adds is what is missing from the payroll number. Unemployment held at 4.1% and the participation rate improved, but "you don't see an acceleration in wages. If you had a heating up labour market, you would expect wages to accelerate, but we don't really see that yet." A labour market adding jobs without paying more is not the classic inflation engine. What he thinks does argue for a hike sits on the other side of the print: "oil and energy prices continue to move higher... especially if you're focusing on refined products like diesel, which hit an all-time high in the US." Reuters and CNBC both carried that diesel record this week, attributing it to Ukrainian and Iranian strikes on refining capacity rather than to demand.
So the hawkish case is an energy case wearing a labour-market headline, and that matters for what settles it. The Fed's target measure is core, not headline. Waller also pointed at a change in how portfolio-management fees enter PCE, due within weeks, which Wang says takes roughly 0.2 percentage points off the Fed's own measure mechanically.
Wang then names a calendar constraint that gets almost no airtime. October is effectively unavailable: hiking then "would be literally a few days before the midterm... too much of a perception of political interference", so "if you skipped September, I think you would have to be thinking of December." If that holds, Friday's CPI is not choosing between September and October. It is choosing between September and December, and a coin-flip contract is pricing a three-month decision.
The August core CPI books lean slightly hotter than they did a week ago. Core CPI year-on-year prices 2.4% as the modal outcome at 47.5%, up 2.5 points on the week, with the 2.5% leg at 21.0% after gaining 8.5 points in a day; core month-on-month prices 0.2% at 53.0%, up 8.5 points on the week. Read the drift, not the level — the modal year-on-year leg carries $1.25k of liquidity and $616 of 24-hour volume, which is noise-level depth.
The consequence is narrow and datable. The August CPI release lands on 11 September, three sessions after the market reopens and three days before the 15–16 September meeting, and on Wang's reading of the calendar it carries a decision about the next three months rather than the next four weeks.
2The surge in global bond yields is an oil story, and the oil story got worse over the weekend
Wang's second segment rejects three of the four popular explanations for the global move in long yields, then makes the one he accepts testable.
On deficits: the US does run about 6% of GDP, but "in the US in particular there are solutions... you just raise taxes", and a top federal bracket of 37% starting near $650,000 leaves room the UK, Canada and Western Europe do not have. He treats fiscal space as a real constraint abroad and a political choice at home. On hyperscalers crowding out Treasuries: "you don't actually have a fixed amount of cash to invest... you can lever up", and the repo market makes Treasuries the hardest asset in the market to crowd out. On strong growth — the explanation John Williams gave CNBC, that higher yields are "really a reflection of the strength of the economy" — Wang is blunt. Growth is 1.5%, was about 2%, "there's nothing particularly strong about that", and the move is global, so a US-growth story cannot carry it. He allows term premium as reasonable given policy and political uncertainty.
What he finds most persuasive is the correlation with the war: "the moment the Iran war began, oil prices began to shoot higher and bond yields have been going higher as well... risk-on, yields up. Risk-off, yields go down." Europe, importing gas, is more exposed to that channel than the US. His falsifiable version is the useful part: if the war resolves, "I would expect bonds to come down, and come down quite a lot", and the fiscal-crisis and hyperscaler-debt stories "will be shown to be quite a trivial part of the recent rise in yields."
The weekend moved that test the wrong way. US Central Command said it struck three Iranian crude oil carriers on Saturday, and Reuters reported US and Iranian forces firing at vessels in waters near Iran the same evening. Polymarket's ceasefire ladder repriced with it: a ceasefire by 11 September fell to 49.5% from 62.0% a day earlier, a 12.5-point drop; by 18 September to 66.5% from 69.5%; by 30 September to 74.5%. Those near legs hold $12k to $54k of liquidity, so the direction is the evidence and the level is not. The already-expired 4 September leg, which traded $141k in 24 hours, fell 6.5 points.
The oil ladder says the same thing with more money behind it. WTI's September high-of-$90 contract has resolved yes. A $95 print prices at 67.0%, up 36 points on the week; $100 at 31.5%, up 21 on the week and 4 on the day; $110 at 9.5%. On the other side, a September low of $80 has collapsed to 33.5% from 92.0% a week ago. Shipping has not recovered at all: Hormuz traffic returning to normal by 15 September prices at 0.5% on $118k of liquidity, and by 31 December at 26.5%, down 4 points on the week.
The cash market is consistent with the framing. The 10-year yield sat at 4.77% on 3 September after printing 4.79% on each of the two prior days, and the 10-year-minus-2-year spread narrowed to 41bp on 4 September from 43bp the day before — the front end selling off harder than the long end.
The consequence is that these first two claims are one mechanism seen from both ends. Wang's duration-relief trade waits on a ceasefire, and the ceasefire book spent the weekend moving away from it. The same escalation is what puts a record diesel print into the headline inflation measure. If his read of the driver is right, a hike in September would be a hike into an oil shock rather than into a wage spiral, and nothing in Friday's core print distinguishes the two.
3The cash flows behind the AI buildout are moving between statements faster than they can be compared
Cameron Dawson, chief investment officer at NewEdge Wealth, spent a stretch of Excess Returns' Clickbait episode on accounting rather than valuation, and her points are checkable claims rather than mood.
extended its accounts receivable to help its customers, she said, while those customers — the hyperscalers — extended their accounts payable. The effect is symmetrical and opposite: it "made Nvidia's free cash flow look worse, but it made the hyperscalers' free cash flow look better." Normalise the payables and hyperscaler free cash flow would read more negative than reported.
lowered its capex guidance, on her account, by reclassifying finance and operating leases so that part of the spend moves out of operating cash flow and into investing cash flow, which meant "the capex calculation that they use looked to be lower than it actually was." She cites a Financial Times piece arguing Microsoft discloses its AI business poorly by bundling it with others, and notes that Microsoft's 25% of OpenAI is carried under the equity method. She expects disclosure around stakes like that to increase over time.
Beyond both statements sit the special-purpose vehicles: "a bunch of stuff that's not even hitting the balance sheet or cash flow statements because it's all happening in these SPVs... super esoteric and very murky."
Her forward claim is dated and mechanical. One-time securities gains are inflating GAAP headline earnings now, and and report on a GAAP basis only. When those gains do not repeat, "as we go into 27, we could be talking about negative year-over-year growth in a quarter like the second quarter" — a comparison problem rather than a demand problem, but one that lands in the printed number. She frames the whole pattern cyclically: "you typically don't see shenanigans as bad, as prolific, as widespread when things are in the earlier stages of an economic boom. You see the most shenanigans at the end."
She is also against the two live proposals to thin corporate disclosure — moving to semi-annual reporting and rolling back post-2009 requirements — on the ground that less disclosure widens the gap between those who can buy management access and those who cannot: "it's who has the resources to be able to gather the information."
The consequence is narrower than a bubble argument. Anyone comparing chipmaker and customer on reported free cash flow is partly comparing a working-capital decision, and both sides moved it in the direction that flatters the customer. That is a disclosure problem before it is a valuation problem, and disclosure problems resolve on someone else's timetable.
4Two Asian regulators pushed back on single-stock leverage in the same week a US filing pushed it further
Dave Nadig set the contrast out on the same episode. South Korea now requires three days of mock trading and five hours of coursework before an investor can buy a 2x single-stock product, imposed after too much money in levered products broke its market for a couple of days — and, he says, the money then left, "which is exactly what you would kind of want to happen." Japan declined single-stock leverage in its own market and asked not to be sold it from abroad. In the same trading week, a US filing landed for 2x leveraged products that reset hourly. "To suggest we have different regulatory approaches here is a massive understatement."
His actual concern sits a layer below leverage ratios. Products are being brought to market whose hedges live in futures that "will not exist until the ETF has assets", the seasoning requirements that once prevented that having gone. The firms who would have to make those markets are the same ones deciding where wheat settles and how rates clear, so the cost is a claim on scarce market-making attention rather than a moral objection.
The demand side has been measured. Nadig cited a Betterment survey finding 26% of Gen Z investors treat sports betting as a deliberate, ongoing part of their long-term financial strategy, and a CNBC figure that 52% of surveyed Gen Z investors had redirected money from investing into sports betting in the preceding year. Dawson's read is that the usual corrective has not fired: an 85% drawdown in ARK's flagship innovation fund did not reset risk appetite the way 2000 did, because every drawdown since has recovered within a year or two, where bought in February 2000 took 14 years to make a new high. She notes high-beta momentum is down 45% from its June peak with sentiment unchanged.
The consequence is that this regulatory risk is not priced as an event, because on Nadig's reading it does not need one. "I think just a change of administration" is enough, or a single state, and the people whose businesses depend on it are, in his phrase, assuming the response has to follow a catastrophe.
5The largest gold fund holds less metal per share every year, by design
Felix, on Felix & Friends, read 's prospectus and made the mechanism the story. The fund charges 0.4% a year and pays that fee by selling gold, so the metal behind each share falls continuously: a tenth of an ounce at the 2004 launch, roughly nine-hundredths of an ounce now. "No one's stealing anything here. The gold just sort of evaporates on schedule, and it says that in writing in the document."
Two further features come from the same document. Shares are exchangeable for metal only by authorised participants — the largest banks — and only in blocks of 100,000, which at Friday's close of $406.78 is about $40.7m. And the custodian may pass gold to sub-custodians down a chain where the fund's own ability to inspect or take legal action is, in the prospectus's own risk section, limited.
His verdict is split, and the split is the point. For price exposure the funds work: "click, click, you get gold price movements cheaply done." For the insurance case — owning gold against a day when institutional trust is tested — he argues you have bought the insurance from the institutions and stored it with them, "a bit like buying a lifeboat that's bolted to the ship." On that reasoning he is out of gold and silver index funds entirely, and he is explicit that this is not a forecast: "I'm not predicting a crash for next Tuesday."
What he holds instead runs in three tiers. Allocated physical metal — serial-numbered bars held off the storage firm's balance sheet — is the foundation, and he notes central banks bought over a thousand tonnes last year in numbered bars rather than fund units. Miners come next, with named as the basket route for anyone unwilling to pick one; their appeal is operating leverage against a largely fixed cost base, and their risk is that same leverage in reverse plus strikes, floods and management. Third is royalty and streaming, which he thinks "quietly beats" the miners: these firms fund a mine up front for a 1–5% revenue royalty or the right to buy production at a fixed discount, so cost blowouts stay the mine's problem. He names as the model's clearest case — about 40 employees, no debt, and keeping over 70 cents of a revenue dollar against a miner's roughly 40 — putting its market value at $50bn, where the current figure is nearer $72bn. His caveats travel with the name: size it small, the model is well enough understood that these rarely trade cheap, and a royalty on a mine that shuts is "a royalty on dust." Gold in total, he says, is an insurance line of 5–15% of a portfolio, "never a bet."
He also offers a paper-market figure as his framing of the gap between metal and claims on metal: COMEX registered gold at about 14.5m ounces, down 23% over the last year, against roughly 40m ounces of open claims. One independence note belongs with all of it — the video routes viewers to a free report and a month's trial of his own stock-screening app, and that app is on screen through the miner and royalty sections.
The consequence is that the two reasons people hold gold come apart. Tracking the price and holding something no counterparty can suspend are different jobs, and on his reading of the document the instrument that does the first well is the one that does the second worst.
What the sources recommended
Three videos produced two single-name calls, both from the same fifteen-minute video, and both are arguments about instrument structure rather than about a price. The other two episodes named no ticker as a call at all — one is a Fed-policy read, the other an accounting and market-structure discussion.
is the buy, and it arrives as the exemplar of a category rather than with a price target: Felix endorses royalty and streaming over mining and names Franco-Nevada as the model's clearest case, while telling viewers to size it small and warning that the model's quality is understood well enough that the shares rarely come cheap. Franco-Nevada's primary listing is Toronto; the line trades in New York.
is the sell, and the reason is the prospectus rather than the gold price. He is explicit that the fund still does the price-tracking job well and that he is not calling a crash. He is out of silver index funds on the same reasoning but named no silver ticker, so none is recorded. appears as the basket route into miners with no view attached to it.
The market gets three sessions before the print that decides the meeting
US markets are closed on Monday 7 September for Labor Day; the next regular session opens at 09:30 ET on Tuesday 8 September. That leaves Tuesday, Wednesday and Thursday before the August CPI release on Friday 11 September, which is three days ahead of the 15–16 September meeting the September contract resolves against. On Wang's calendar argument, that one print is carrying the September-or-December question rather than a four-week one.
The earnings in the window are second-tier for the macro question but not empty: , , , and on the 8th; , , , , , and on the 9th; and after the close on the 10th.
The variable with no calendar entry is the one that moved most this weekend. A Saturday strike on three tankers took 12.5 points off a ceasefire contract while the equity market was shut, and on the reading in claim two that is the same input that decides where the 10-year opens on Tuesday.
