The energy contribution to inflation is easing while core stays sticky, so cheaper oil no longer buys a rate cut. The 30-year Treasury clears 5 percent, and leadership keeps leaving the AI mega-caps for small caps and emerging markets.
Shares at a high, oil firm, volatility asleep, gold slipping, and the ten-year yield creeping up. Four of these five are calm. The fifth is the story, and its longer-dated cousin, the thirty-year, has just cleared 5 percent.
“Inflation rarely leaves. It changes clothes, and the market keeps watching the door it came in by.”Anthony Rosenthal, Weekly Market Pulse, July 2026
TL;DR
Markets in brief. The S&P 500 closed Friday 10 July at 7,575, up about 1.2 percent on the week and 10.7 percent for the year, a fresh high. The 10-year Treasury yield, which sets the cost of mortgages and business loans, sat at 4.54 percent and the 30-year closed at 5.05 percent; the two-year held near 4.16 percent. WTI crude firmed to 71.51 dollars on renewed Iran headlines. Gold slipped to about 4,129 dollars as real yields rose. The VIX, Wall Street’s fear gauge, held near 15, its calmest range of the year.
The one red light on the dashboard is the same as last week, a normal-shaped yield curve, and the calm everywhere else is exactly what makes the rate move underneath it easy to miss.
An extended hold is the base case; the hike the market is already pricing is the tail we respect but fade, and a later cut is the smaller tail the labour rollover keeps alive.
| Scenario | Probability | Trigger | 2Y | 10Y | Equity impact |
|---|---|---|---|---|---|
| Extended hold, core sticky (base) | 55% | Core CPI firm (≥0.3% MoM); labour soft but not breaking | ~4.15% | ~4.55% | No cuts in 2026; the long end stays heavy; leadership keeps rotating to value, small caps and EM |
| Tariff/core reflation delivers the hike | 30% | A hot core print plus tariff/rare-earth goods pass-through re-accelerates | >4.3% | >4.7% | The September hike the market is already pricing lands; a genuine risk-off |
| Disinflation resumes, a later cut reopens | 15% | June core CPI soft (≤0.2%); the labour rollover deepens | <4.0% | <4.4% | The cut path reopens later in the year; duration rallies, the broadening accelerates |
Reconciliation: the market itself, via futures, prices roughly two-thirds odds of a September hike. Our base case is a hold, so we are deliberately fading the market, and we should say so plainly. Three practitioner desks we track (Slok at Apollo, Neuberger Berman, Deutsche Bank Wealth) read an extended hold with hike odds nearer a quarter than two-thirds, and the June labour rollover keeps a later cut, not a hike, as the genuine tail. We hold a 30 percent hike probability out of respect for the market’s pricing, not agreement with it.
How the 15 is built: each signal scores 0 when green, 5 when amber and 10 when red; the scores are weighted and summed. The weights, the thresholds and this week’s full working are printed in the appendix, A6, so you can recompute the number yourself.
Last week’s tell, scored first. The tell was specific: Brussels had until Saturday to ratify the trade deal, and if it ratified and September hike odds fell back below 40 percent by Friday, the tariff tail would close and the disinflation handoff would run clean. Half happened. Brussels ratified, industrial-goods duties came off from 1 July, and the deal landed at the 15 percent cap, so the tariff tail did close. But September hike odds did not fall below 40 percent; they rose, to roughly two-thirds on futures, as the New York Fed’s one-year inflation expectations hit a three-year high and core PCE printed 3.4 percent. The handoff did not run clean. The composition of the risk simply rotated, from tariffs to core, and that rotation is this week’s argument.
The interest rate on a thirty-year US government loan closed at 5.05 percent this week, its highest of the cycle. Here is the distinction that matters, because it is easy to get wrong: crude oil is actually up 13 percent on the year and firmed to 71.51 dollars this week on Iran headlines, so the falling-energy story is not about the barrel price. It is about what the bond market expects inflation to be, market-implied expectations easing to roughly 2.2 percent, a twelve-month low, even as the cost of money rises. Those two are supposed to move together. This week they came apart. Apollo’s Torsten Slok gave the cleanest name for it: when oil and Treasury yields decouple, it signals that the thing binding the Fed’s hands has moved from headline inflation, the numbers a household feels first, to core inflation, services and rents and the slice of goods now carrying an embedded tariff.
Here is why a saver or a borrower should care. A homeowner waiting to remortgage, a business waiting to refinance, a government funding its deficit at the long end: all of them were quietly hoping the falling oil price was the first domino toward cheaper money. It is not, and the reason is mechanical. Core inflation is made of wages and services, and wages do not fall the way a barrel of oil does. The good news at the petrol pump is real. It is simply not the news that sets your mortgage rate anymore.
The Fed Path read, and why it moved. Last week our internal Fed Path Model leaned toward a cut arriving on soft labour data. This week it reads Hold, with a hawkish tilt we will not confirm until Tuesday’s CPI. Be precise about the discipline, because it is the whole point of the model. Two of the three blocks now agree on hold, do not cut: the inflation block (core PCE 3.4 percent, one-year expectations at a three-year high of 3.7 percent) and the market block (the 30-year above five, the Taylor Rule prescription still well above the funds rate). That two-block agreement is what moves the read from cut-lean to hold. But only one block, inflation, argues for an actual hawkish tilt; the labour block is squarely dovish (June payrolls a soft 57,000, prior months revised down by a net 74,000). By the model’s own two-confirmation rule, one hawkish block is not enough to lean hawkish, so the tilt stays conditional on the 14 July print, and we flag the timing openly: leaning hawkish two days before a CPI release is the exact mistake this model was built to avoid.
The broadening, explained mechanically. Look down our own Scoreboard and the story is not that the giants have fallen. It is that everyone else has finally caught them. Small caps are up 20.0 percent this year and emerging markets 19.0, and both have now drawn level with the Nasdaq 100 at 18.4, after two years of being left behind by it. The broad S&P, at 10.7, trails all three. That is not the market breaking, it is the market broadening: not a collapse at the top, but the rest of the market drawing level with it. When money leaves three enormous crowded names and spreads into two thousand smaller ones, the index can keep rising while its most famous members fall, and breadth rises rather than falls. Think of an orchestra where the three loudest soloists step back and the rest of the players come up in the mix: the piece gets fuller, not quieter.
Competing-driver discipline. It would be self-serving to read the whole rotation as confirmation of our Year-of-the-Repricing thesis. The honest competing explanation is more mundane: a lot of this is positioning, and the mega-cap AI trade was the most crowded position in the market, which unwinds for its own reasons regardless of the macro story. The tell that separates the two: if this is a genuine repricing, the money leaving the AI names keeps flowing into cheaper, never-crowded corners even after the crowding clears; if it is just a positioning unwind, the flows stall once the AI names find their sellers and leadership snaps back. Watch where the money goes after the AI selling exhausts, not during it.
Where I could be wrong (the standing self-check): First, the hawkish-hold read leans on core staying sticky; if the 14 July core CPI prints a soft 0.2 percent, the labour block’s dovish vote wins and the cut path reopens. The tell: a sub-0.3 percent core print and September hike odds falling back through 50 percent. Second, the constructive read on the broadening may be too sanguine; Michael Howell’s liquidity work argues the global tide is going out from a late-2025 peak and the 2027 refinancing wall is the real danger, which would break the broadening from underneath. The tell: high-yield spreads widening through 350 basis points, from 270, while the index is still near highs.
30Y closed 5.05%; core PCE 3.4%; NY Fed 1-year inflation expectations 3.7% (3-year high); June payrolls 57k, revisions -74k; breadth 67.2%; small caps +20.0% and emerging markets +19.0% have drawn level with the Nasdaq 100 at +18.4% while the broad S&P sits at +10.7%; the crash gauge at 15.
The base case is an extended hold with no cuts in 2026, and we are fading the market’s two-thirds odds on a September hike rather than agreeing with them. The leadership rotation is part genuine repricing and part positioning unwind, and we cannot yet tell which dominates.
A soft 14 July core CPI; a second weak labour print flipping the July “live meeting” read; a credit-spread widening that turns broadening into breaking.
The founder, in one word: the Historical Renovator. David Reger does not invent new markets so much as walk into broken ones, work out why everyone before him failed, and retrofit the missing piece.
Who he is. Reger did not start in a robotics lab. He started as a technical model-maker, a maker of the precise physical prototypes engineers build before they trust a design, and in 2009 he did what a certain kind of ambitious European does: he went to Silicon Valley, where he spent part of his time working as a community social worker. He came back to Europe in 2013 convinced of an unfashionable idea, that robots had failed in homes, shops and small factories not because they were too expensive, but because they were too stupid, blind and deaf and unable to feel, and that the whole industry had been solving the wrong problem. He built two robotics companies to test the thesis, then founded Neura in 2019 to bet everything on it. This summer the bet got its verdict: a Series C of up to 1.2 billion euros, roughly 1.4 billion dollars, led by the stablecoin issuer Tether at a valuation near 7 billion dollars, with Nvidia, Amazon, Qualcomm, Bosch and Schaeffler alongside. Neura calls it the largest round ever raised by a full-stack robotics company; at roughly 1.4 billion dollars it is, on any measure, among the largest a German company has ever raised (Bloomberg, June 2026).
The problem. For thirty years the robotics industry sold the same thing: a machine that repeats one motion perfectly in a cage, bolted to a factory floor, blind to anything outside its programmed arc. It works brilliantly for a car assembly line and is useless everywhere else, because everywhere else is unpredictable. Reger’s diagnosis was that the bottleneck was never the arm or the motor, which are cheap and getting cheaper; it was the integration of sensing, cognition and action into one machine that could handle a world it had not been pre-programmed for. Fix the thinking, he argued, and the same hardware suddenly works in a warehouse, a workshop, a hospital corridor.
The moment of real uncertainty. The uncertainty is on the record and it is enormous. Neura is not selling a box, it is trying to own an ecosystem, a platform it calls the Neuraverse that other developers plug into, and it has staked that ambition on two industrial partnerships that will either make it the backbone of European factory robotics or leave it a well-funded also-ran. The Schaeffler deal commits the two firms to develop actuators together and to deploy a mid-four-digit number of Neura humanoids across Schaeffler’s global plants by 2035, a roughly 300 million euro programme; the Bosch partnership is meant to industrialise production. Both are bets that the cognition claims convert into reliable, cost-effective uptime on a real factory floor, day after day, and that is precisely the constraint that has humbled every humanoid demo before it. If the robots cannot hold reliability at plant scale, the platform story collapses into a hardware story, and hardware stories do not carry 7 billion dollar valuations.
What it teaches. The lesson is not that robots are the future. It is Reger’s diagnostic discipline: he found the bottleneck everyone else had priced as a cost problem, make the arm cheaper, and re-read it as a systems problem, make the machine understand, then built the thing others would have to plug into rather than the thing others would compete with. The transferable move, in any industry, is to ask whether the constraint everyone is racing to make cheaper is actually the constraint at all.
The wave it is riding, named. Neura is not alone, and that is the point. On the Thursday screen’s figures (TechCrunch, Forbes and PitchBook, July 2026), robotics startups have raised roughly 18.8 billion dollars globally in the first half of 2026, against about 15 billion across the whole of 2025. The names tell the story: Figure AI at a reported 39 billion dollar valuation, AI2 Robotics in Shenzhen near 3 billion, and, most tellingly for public-market readers, Agility Robotics filing to go public via a 2.5 billion dollar SPAC, which its backers describe as the first pure-play humanoid to reach the stock market. The same wave funding Neura privately is now washing up onto public markets through Agility, which means retail investors are about to get a priced, tradeable test of the exact labour-substitution economics this Case Study is about. We are holding Agility for On the Radar until it trades. When it does, the number that will matter is the gap between its listing price and what the private market says its peers are worth.
The two-sided risk. First, labour. Neura’s own flagship deal is a plan to put thousands of humanoids into factories by 2035; plant-scale deployment of cognitive robots is, by design, a substitution of capital for labour in exactly the semi-skilled industrial jobs that anchor communities. The technology’s promise, safer and more productive plants, and its social cost, who does the work and who captures the gain, are the same event seen from two sides. Second, governance. The round is led by Tether, a stablecoin issuer whose reserves, disclosures and regulatory standing remain contested. A lead investor’s balance sheet becomes part of a company’s risk surface, and a champion of European strategic autonomy anchored by a lightly-regulated offshore crypto issuer is a genuine question, not a footnote.
Anthony Rosenthal analysis score: 6.2 out of 10, WATCHLIST, INTEGRITY AMBER. The framework validates a genuinely differentiated thesis and cap table, priced ahead of proven reliability at scale; the named zero-scenario is that cognition fails to convert to reliable plant-scale uptime and the platform story de-rates to a hardware one.
The index rose while its generals sat it out. The tape did the quiet, awkward thing it has done all year: it went up at the index level while its most famous constituents went sideways. The S&P 500 closed at 7,575, up on the week and a fresh high, carrying its year-to-date gain to plus 10.7 percent. Beneath it the small-cap Russell 2000 (plus 20.0 percent for the year) and emerging markets (MSCI EM plus 19.0 percent) kept outrunning the mega-caps, the clearest single sign that leadership is changing hands.
Commodities split by story. WTI crude firmed to 71.51 dollars (plus 13.2 percent for the year) on renewed Iran headlines, recovering from the early-July sub-68-dollar low without a confirmed new supply outage, a de-escalation-premium wobble rather than a supply shock. Copper held firm (plus 10.6 percent) while silver had an ugly week, down about 4 percent to 60.30 dollars (minus 14.6 percent for the year), and the Baltic Dry Index, the cost of moving raw materials by sea and a real-economy pulse-check, jumped again to 2,944 (plus 56.4 percent for the year, and up more than 8 percent on the week alone), still the year’s best line on the board.
The bond side is where the meaning sits. Long-dated Treasuries fell again (TLT minus 10.4 percent for the year), the aggregate bond index stayed underwater (AGG minus 3.98 percent), and gold, usually the beneficiary when real yields misbehave, instead slipped to 4,129 dollars (minus 4.89 percent), a reminder that a rising real yield, the return after inflation, is a headwind for a metal that pays no coupon.
The froth is leaving from the edges. Crypto stayed in its own bear, Bitcoin minus 27.2 percent and Ethereum minus 39.5 percent for the year, the two weakest lines on the Scoreboard, and that is worth pausing on. It is fashionable to call this a calm market, and on the visible dials it is. But the most speculative corner, crypto, and the second most, silver, are quietly draining even as equities make new highs. That is exactly the pattern Michael Howell’s liquidity work would predict if the global tide is going out: the froth leaves first, from the edges, long before it reaches the core. The internals are not uniformly calm, and that is the honest footnote on a crash gauge reading of 15.
Earnings season opens. The season kicked off with steady staples, PepsiCo affirmed guidance on organic growth of about 2.4 percent, and the banks and Delta reporting into the weekend. The real test of the broadening is whether the average company’s earnings justify the average stock’s new leadership; the next two weeks answer it.
The dashboard is mostly green, and the one thing it is not built to see, the acceleration of the credit financing the AI build-out, is exactly where this week’s risk is hiding.
The standing qualifier. One analyst keeps every green light above honest. Michael Howell, whose liquidity-cycle work this publication tracks, reads global liquidity as topping from a late-2025 peak, with roughly 40 trillion dollars of debt to refinance by 2027 into a pool that has stopped growing. If he is right, today’s tight spreads describe the weather rather than the climate, and that refinancing wall is the named pressure point behind every benign reading above.
A composite score of 15 out of 100 says the same thing in plain language: on the visible dials this is a calm market, unchanged from last week and the joint-calmest reading of the year so far, with only the shape of the yield curve flashing red and everything in credit, volatility and breadth reading green. What the score is not built to measure is the one risk this week’s research keeps circling, the financing behind the AI build-out. The dials tell you the house is not on fire. They do not tell you whether the mortgage on it can be refinanced. That is the Contrarian Corner’s job below. Position for normal volatility; keep the hedges that pay off if the curve steepens further.
The number that should stop you this week is 15.8 percent. That is the share of real, end-to-end freelance projects, computer-aided design, video editing, data analysis, building a working web app, that a frontier AI model can now complete well enough to be paid for, measured by the Remote Labor Index built by the Center for AI Safety and Scale. In October 2025, when the index launched, the best model managed 2.5 percent. In under eight months the frontier has more than quadrupled. This is not a benchmark of trivia or exam questions; it is a measure of whole jobs done end to end, and it is the single most concrete piece of evidence this week for the side of the AI-labour debate that says substitution is accelerating.
This takes about four minutes. Paste the following into Claude or your model of choice: “Take one real task from my job that takes me about two hours, [describe it in one sentence]. Break it into the sub-steps you could do end-to-end today with no supervision, the sub-steps you could draft but I would need to check, and the sub-steps you genuinely cannot do. Be honest about the middle category.”
What Anthony found when he ran it: the honest answer is almost never “all of it” or “none of it,” it is a precise map of the middle, the drafting you can hand over versus the judgement you cannot. The transferable insight is that the Remote Labor Index’s 16 percent is an average hiding a bimodal reality: some of your work is already at 90 percent, some is at zero, and the skill worth building this year is knowing which of your tasks sit in which bucket, because that is where your time is about to be revalued. Run it on your own role before someone else does.
The tariff weapon turns into a trade-flow weapon, and that leads this week, not Hormuz. The most consequential geopolitical development for portfolios this week was not a missile, it was a threat about routing. Washington’s proposed 25 percent tariff on the trading partners of Iran, layered on top of an EU deal that landed at a 15 percent cap, keeps a live overhang on global trade flows, and it lands in the same month that China’s grip on rare-earth and critical-mineral processing, roughly 85 percent of refined rare earths, remains the West’s unresolved chokepoint. The mechanism a reader should hold: tariffs used to be about the price of a good; increasingly they are about whether it can move at all, and a supply chain that can be switched off at a border is a different kind of risk than one that is merely taxed. This is the channel most likely to reintroduce goods inflation if the “core goods at zero” reading, tariff pass-through supposedly complete, turns out to be premature.
Hormuz, signed and already unravelling, and we owe you a correction. On 17 June the United States and Iran signed a fourteen-point memorandum in Islamabad: the Strait would reopen, Iranian crude would flow under Treasury waivers, and technical talks on the nuclear programme would follow. We have described that memorandum as unsigned in recent editions, and that was wrong. Three weeks on it is coming apart anyway. On 7 July, after renewed Iranian attacks on shipping in the Strait, Washington revoked the waivers. That is the whole of this week’s oil bid: crude firmed to 71.51 dollars on the headlines with no confirmed new supply outage, and tankers have continued transiting at roughly three-quarters of prewar export levels. The supply-fear reflex is still not firing. But the lesson has moved. We said for nineteen weeks that a ceasefire is not peace. The sequel, learned the hard way: a signature is not compliance. The residual risk remains the slow one, ballistic-missile and production reconstitution, and force-majeure insurance clauses stay the leading indicator to watch for genuine re-escalation.
The parallel worth naming, because it is happening to us in both directions. This publication’s worst-performing call and its best-performing call are now being hit by the same chokepoint from opposite ends. MP Materials is down 22.3 percent since we backed it, because Beijing named it directly in its rare-earth retaliation. And this week Bloom Energy, our strongest call of the half, was attacked by a short-seller alleging the opposite exposure: that a company built on the promise of an American, China-free supply chain is in fact quietly dependent on Chinese scandium. One call was punished for being outside the chokepoint. The other is being punished for the suspicion that it is secretly inside it. That is what a genuine chokepoint does. It does not just raise the price of a material; it makes the provenance of every supply chain a matter for the equity analyst, and it turns “where does this actually come from?” into a question a company can be shorted on. Both calls are scored in Portfolio Watch below, and neither is being quietly dropped.
The developed-market bond divergence. The 30-year US Treasury above 5 percent is not happening in isolation; long-end yields are under pressure across developed sovereigns as term premium, the extra yield investors demand for lending long, rebuilds globally. The read-across is that the “duration is cheap again” call needs the term-premium regime to stop rebuilding before it pays, which is precisely the second watch condition in the Analytical Takeaway.
The contrarian claim. Everyone is arguing about whether AI revenue is big enough. That is the wrong question. The number that decides whether the build-out is safe is not the level of revenue or even its growth rate, it is the acceleration, the second derivative, and that is a far more fragile thing to depend on.
Across the four big AI platforms sits roughly 2.1 trillion dollars of what accountants call remaining performance obligations, contracts signed but not yet delivered (Groundbreaker, “The Second Derivative,” 2 July, aggregating the platforms’ disclosed backlogs). That backlog is regularly cited as the bull case: proof the demand is real and locked in. Here is the whole argument in one sentence, before any of the machinery: the AI build-out is being paid for with borrowed money, and the loans only work for as long as growth keeps speeding up. Read the backlog less as a pile of future sales and more as a giant loan book extended to a handful of cash-burning frontier labs, and the risk picture inverts. The lending arrives in unfamiliar clothing: take-or-pay compute deals, where you pay for the capacity whether or not you use it, like a gym membership you cannot cancel; term loans secured against the chips themselves; off-balance-sheet vehicles, financing companies keep off their main accounts; and bundles of these contracts sold on to insurers as asset-backed securities, a bond whose repayments come from one specific stream of promises. The whole structure stays current only while revenue keeps accelerating. When growth merely slows, revenue can still be at a record high while the financing that assumed faster growth quietly breaks. Groundbreaker’s phrase for it is “borrowed time”: record revenue, machine already broken.
The analogy that makes it click. This is 2008 mechanics, not 2000 mechanics. The dot-com bust was a slow de-rating: valuations were too high, and they came down over months as the story faded. A credit-driven real-estate cycle is different; it does not de-rate slowly, it seizes, because the whole structure depends on being able to refinance, and the day refinancing is unavailable the asset changes hands overnight. If AI’s build-out is financed like property rather than valued like software, the failure mode is a funding stop, not a fading multiple. Slok’s companion point sharpens it: market-implied ratings on software leveraged loans already show far more credit dispersion than the agencies’ ratings admit, the market is quietly pricing risk the ratings have not caught up to.
About 2.1 trillion dollars of contracted-but-undelivered backlog across the big four; the build-out is increasingly debt-financed (take-or-pay, chip-collateralised loans, asset-backed notes sold to insurers); market-implied software-loan ratings show more dispersion than agency ratings.
The binding risk is a deceleration of bookings growth, not an absolute fall; the failure mode rhymes with 2008 (refinance-or-seize), not 2000 (slow de-rate); insurers now hold a slice of the tail.
A genuine capability step-change that re-accelerates enterprise adoption (bullish, hands it back to the Augmenter); or the first visible crack, a delayed hyperscaler payment, a pulled term loan, an asset-backed tranche struggling to place.
Discipline note. This is a framework for where to look, not a crash call. The crash gauge is 15 precisely because none of these cracks are visible yet. The point of a Contrarian Corner is to name the seam before it shows, not to predict the day it tears.
In April this publication told you copper was the cleanest way to own the artificial-intelligence build-out, and that the way to own copper was Freeport-McMoRan. Copper is up 10.6 percent this year. Freeport is down 6.1 percent since we said so. We were right about the metal and wrong about the miner, and the gap between those two things is the most instructive mistake of our first half.
The full account, before the excuses. Nineteen four-week checkpoints have been scored since Week 14. Thirteen were right and six were wrong, a hit rate of about 68 percent. Nine calls now carry a thesis horizon under the dual-horizon rule and not one has closed; the first two verdicts, Bloom Energy and Freeport, land on 18 July. Of the nine live calls, four are in the money (Arista, Bloom Energy, Talen and Mitsubishi UFJ) and five are behind. The best call of the half was Bloom Energy, up 46.6 percent at its four-week mark. The worst is MP Materials, down 22.3 percent from entry.
Three of the six losses are the same mistake. Freeport (down 4.1 percent at its checkpoint), MP Materials (down 2.6 percent then, down 22.3 percent now) and Venture Global (down 6.9 percent) were all equity proxies for a commodity view. Here is the uncomfortable part: the commodity views were excellent. Our three crude oil calls scored 91, 90 and 80 out of 100. The oil was right nearly every time. The companies that pull it out of the ground were not.
The mechanism is not mysterious once you say it out loud. A commodity price is one variable. A mining company is that variable plus a balance sheet, a cost curve, a jurisdiction, a capital-spending cycle and, as MP Materials discovered when Beijing named it directly, a geopolitical target on its back. Every one of those is a way to be right about the thesis and lose money anyway. So what, and you can use this: if you have a view on a commodity, the cheapest way to be right is usually to own the commodity. We have added a standing rule. When a call is a commodity view wearing an equity costume, we now have to say so at entry and justify why the equity is the better expression.
We turned bearish on Bitcoin at roughly 77,800 dollars. Four weeks later it was slightly higher and the call scored as a miss. Bitcoin is now down 27.2 percent on the year and is the second-worst line on our own Scoreboard. The analysis was right. The clock was wrong. That single experience is why the dual-horizon system exists: every call is now scored twice, once at four weeks to test the timing and once at a declared thesis horizon to test the analysis, and a call that is early is reported as early rather than quietly reclassified as right later.
On 10 June we hardened the rate view on the strength of a single jobs number and the market’s own hike pricing. Two days later core inflation came in soft and the view looked foolish. That is the reason the Fed Path Model now exists, with its two-confirmation rule, and it is precisely why this week we refuse to confirm a hawkish tilt before Tuesday’s inflation number, even though the inflation block is the loudest thing on the panel. The lesson: never let a single print, or the market’s own hike pricing, move the rate view on its own.
The oil calls, repeatedly. The wall-of-worry read that kept us from turning bearish on equities into the spring, which the track record shows is the direction our errors have always run. Bloom Energy and Arista, the two best calls on the book, both of them bets that the shortage in the AI build-out was physical, power and networking, rather than intellectual. And the standing refusal, across nineteen weeks, to treat any Strait of Hormuz headline as resolved on rhetoric alone, which kept us out of every false supply-fear rally. That discipline earns its own correction this week: the memorandum was in fact signed on 17 June, and the waivers it created were revoked on 7 July after fresh attacks on shipping. The rule survives the error and is sharper for it. A ceasefire is not peace, and a signature is not compliance.
The honest summary. Roughly two-thirds of our four-week calls were right, and the third that were wrong were wrong for a reason that is now written into the process rather than apologised for in a footnote. The verdicts that actually matter, the thesis-horizon ones, have not started landing yet. The first two arrive in eight days, and one of them, Bloom Energy, has just been hit by a short-seller report. We will report it either way.
Each week the WMP scores the central economic argument about artificial intelligence as a contest between two forces. The Displacer is the case that AI substitutes for human labour, concentrating the gains in capital and eventually destroying the spending power the economy runs on. The Augmenter is the case that AI raises human productivity, expands output, and spreads the gains broadly. Both sides agree AI is powerful. They disagree about whether it is a demand shock or a supply shock.
A return to the 3-1 pattern after last week’s draw. Four pillars, one point each:
Week 26: the Augmenter 3, the Displacer 1. Running total, the Augmenter 21, the Displacer 7. What would flip the score: a named white-collar layoff wave citing AI as the cause would hand the Displacer Pillar 2 next week; a fall in the 30-year that cheapens the build-out’s financing would harden the Augmenter’s compute-ceiling point further.
The ERDR (Equity Return for Debt Risk) framework tracks twelve income strategies that aim to earn an equity-like return for taking on debt-like risk. This is an even week, so all twelve are refreshed in the Standing Dashboard without a Deep Dive. The terms, briefly: a spread is the extra yield a strategy pays over a safe government bond, in basis points (hundredths of a percentage point); investment grade means the safest tier of corporate borrowers; and a floating-rate strategy is one whose income rises automatically when short-term interest rates rise.
With the 30-year above 5 percent and high-yield spreads tight at 270 basis points, the spread strategies are being paid less for the same credit risk while the duration-heavy income strategies carry mark-to-market pressure from the long end.
| Strategy | Indicative Yield | Spread vs IG | WoW | Action |
|---|---|---|---|---|
| 1. Active income fund + Lombard | 6.9 to 7.7% | +300 to 380bps | Funding cost steady; front end little changed | Watch |
| 2. IG / split-rated CLO tranches | 6.3% | +220 to 260bps | Floating coupon drifts with rates | Hold |
| 3. Listed infrastructure debt/equity | 5.7% | +175bps | Discounts widen as the long end backs up above 5 percent | Watch |
| 4. Business development companies | 10.8% | +560bps | Floating yields firm; watch credit quality | Hold |
| 5. Agency mortgage REITs | 13.2% | +150bps (asset OAS) | Hurt by the long-end backup | Watch |
| 6. Senior secured leveraged loans | 8.5% | +420bps | Floating-rate, resilient; but see the rating-dispersion flag | Watch |
| 7. Preferred shares / hybrids | 7.3% | +295bps | Duration drag as the 30-year cleared 5 percent | Hold |
| 8. Real asset royalties | 6.6% | +255bps | Steady; oil firm near 71 dollars | Hold |
| 9. EM hard-currency sovereign carry | 7.9% | +385bps | Carry intact; EM tailwind as the broadening favours it | Hold |
| 10. High-yield municipal bonds | 6.2% (tax-free) | +275bps | Softer as long rates backed up | Hold |
| 11. Private credit direct lending | 11.0% | +575bps | Spreads holding; rating-dispersion and liquidity-turn caution | Watch |
| 12. Trade & supply-chain finance | 8.9% | +450bps | Short-tenor, defensive | Add |
Each strategy is explained in full when it is the week’s Deep Dive.
The read this week: with the long end backing up above 5 percent, the duration-heavy strategies, listed infrastructure, preferred and hybrid capital, high-yield munis, carry mark-to-market pressure and their notes are cautious. With high-yield spreads this tight, the carry strategies are being paid for calm that is already priced. The standing caution is the leveraged-loan and private-credit complex (strategies 6 and 11), where Slok’s chart shows market-implied ratings pricing far more dispersion than agency ratings admit, the exact seam Contrarian Corner probes. No thesis changed this week; the moves were rates, not risk.
What I am watching and why, not a recommendation to buy or sell. One name returns to the main section this week on a genuine new catalyst; every active call is tracked to its score date in Portfolio Watch below.
The ledger first. The running tactical scorecard stands at thirteen of nineteen four-week checkpoints correct, about 68 percent, counting every tactical checkpoint logged in the call record since Week 14; nine calls now carry a thesis horizon and not one has closed, with the first two verdicts due 18 July: Bloom Energy, now fighting a short-seller report over its scandium supply chain days before its verdict, and Freeport, which is behind. The full first-half account is in the Mid-Year Reckoning above. No call is closed until its thesis date, and the laggards are named with the winners: MP Materials, down 22.3 percent since entry, remains the loudest loser and is named as such in Portfolio Watch below, alongside USA Rare Earth and Freeport; Venture Global and YPF both recovered this week as crude firmed.
Arista Networks (NYSE: ANET), re-promoted on a genuine new catalyst. Arista qualifies to return to the main section because it cleared the strict new-catalyst test. On 8 July the company told the market that demand for its 1.6-terabit 7060XE7 platform, announced back on 9 June, is running ahead of plan as it begins supplying it, and named Meta, Microsoft and Oracle among the first deployments. The stock rose about 7 percent that day. That is a named, company-specific event inside the seven-day window, not a price drift. The honest caveat, stated up front: the switches do not reach general availability until the fourth quarter, so the market is paying today for revenue that arrives next year.
What the market may be missing. The AI-infrastructure narrative has been almost entirely a compute and power story; the networking layer that stitches tens of thousands of accelerators into a single training fabric is a quieter, higher-margin choke-point, and Arista has the category’s fastest platform with the three biggest buyers already named on it. As leadership rotates away from the crowded mega-cap AI names, a pick-and-shovel supplier with a fresh product cycle is a different exposure than the hyperscalers themselves.
The historical parallel. Cisco in the late-1990s internet build-out, the switch vendor that compounded quietly while the attention, and eventually the blow-up, sat with the dot-coms it connected. Stated with the cautionary half attached: Cisco’s own multiple eventually ran far ahead of its growth, and the networking supplier is not immune to the same de-rating it enables.
What would change the thesis. A hyperscaler in-housing its switching, the white-box and merchant-silicon threat, at scale, or an AI-capex deceleration that hits the networking order book.
Horizon and the four-week driver, declared. The declared horizon runs more than three months from the Week 20 entry, to 28 August, so the dominant near-term price driver is declared: Nasdaq-100 and AI-capex beta, currently SUPPORTIVE, the switch launch landed into a firm tape. Entry logged Week 20 at 147.00 dollars; the 10 July close was 186.96 dollars, plus 27.2 percent, having jumped on the demand statement. The four-week tactical was already scored correct back on 26 June (plus 12.6 percent); the thesis-horizon verdict falls on 28 August. Benchmark: the Nasdaq 100.
Arista Networks (NYSE: ANET) · entry $147.00 (Wk 20) · close $186.96 (10 Jul) · +27.2% · 4-week checkpoint scored 26 Jun ✓ · thesis horizon 28 Aug · benchmark Nasdaq 100
Also watched, not yet a call: Agility Robotics, the first pure-play humanoid heading for public markets via a cash-shell merger agreed on 24 June. It is covered in the Case Study above and stays there until it trades with a verified price and the full pre-screen can run.
Spare a thought for the yield curve, the most over-interpreted line in finance. For two years its inversion, short rates above long rates, was the recession klaxon every commentator pointed at. This year it quietly re-steepened back to normal, long rates comfortably above short, and the same commentators had to decide whether the un-inversion was the reassuring all-clear or the actual recession signal, because, inconveniently, historically it is often the un-inversion that precedes the downturn. The curve, in other words, is a horoscope that is right often enough to keep its readers, and vague enough to never be sued. This week it did the quietly consequential thing instead: it steepened again, from plus 31 basis points to plus 38, while everybody watched the thirty-year. The signal nobody was reading moved. The one everybody was reading did not.
The pump price kept sliding downhill,
While the long bond broke five with a chill.
Cheap oil, it turns out,
Buys nothing devout,
When it’s core that is footing the bill.
Three things to watch next week, and they are all really one question wearing different hats. Tuesday’s core inflation print, whether the thirty-year holds above five for a third and fourth session, and whether the broadening survives contact with earnings season, which begins with the banks. If all three point the same way, core firm, long end heavy, leadership still rotating, then the extended-hold-and-broaden regime is confirmed and the year’s strangest feature, an index at highs led by everything except its former generals, becomes the settled shape of the second half. If they diverge, a soft core print with a heavy long end, or a broadening that stalls the moment a bank misses, then we are not in a new regime at all, just a crowded trade unwinding, and the old leadership is one good earnings week from taking the baton back. The bond market has already cast its vote. Earnings season is the recount.
2026 Scoreboard
25 assets ranked by year-to-date return · Baselines locked 1 January 2026 · Close of Friday 10 July 2026
The 25 are a fixed basket, set on 1 January and unchangeable during the year: eleven equity indices, four bond and credit funds, six commodities, two currencies and two cryptocurrencies. Every row below is one of those 25, and nothing is added, dropped or substituted mid-year, which is the only way a year-to-date table means anything. The single-company calls in Portfolio Watch are tracked separately and are not part of this basket.
Shipping rates, the Turkish lira and Japanese shares lead the year; crypto and long bonds still hold the floor. A 96-point spread between best and worst, and a top five without a single US mega-cap. But the thesis we registered in January is measured on five asset classes, not twenty-five, and on that measure it is only weakly confirmed so far (see the method note). We report the test we set, not the flattering one. This is the Repricing thesis in a single row of bars.
2026 YTD Performance, All 25 Assets, Week 26
| Rank | Asset | 1 Jan 2026 Baseline | Week 26 Close | YTD % |
|---|---|---|---|---|
| 1 | Baltic Dry Index | 1,882 | 2,944 | +56.43% |
| 2 | USD/TRY | 35.40 | 46.98 | +32.70% |
| 3 | Nikkei 225 | 51,830 | 68,557.73 | +32.27% |
| 4 | Russell 2000 | 2,481.91 | 2,977.81 | +19.98% |
| 5 | MSCI EM | 1,595.20 | 1,897.75 | +18.97% |
| 6 | Nasdaq 100 | 25,200.50 | 29,825.11 | +18.35% |
| 7 | WTI Crude | $63.20 | $71.51 | +13.15% |
| 8 | S&P 500 | 6,845.50 | 7,575.39 | +10.66% |
| 9 | Copper | $5.682 | $6.285 | +10.61% |
| 10 | Euro Stoxx 50 | 5,740.15 | 6,269.97 | +9.23% |
| 11 | Swiss SMI | 13,248.10 | 14,235.09 | +7.45% |
| 12 | FTSE 100 | 9,948.30 | 10,497.29 | +5.52% |
| 13 | DAX | 24,540.20 | 25,067.09 | +2.15% |
| 14 | HYG | $78.15 | $79.71 | +2.00% |
| 15 | Nifty 50 | 24,420 | 24,206.90 | -0.87% |
| 16 | LQD | $109.02 | $107.46 | -1.43% |
| 17 | AGG | $102.15 | $98.08 | -3.98% |
| 18 | Gold | $4,341.10 | $4,128.90 | -4.89% |
| 19 | USD/ZAR | 17.55 | 16.346 | -6.86% |
| 20 | Hang Seng | 26,340 | 24,175.12 | -8.22% |
| 21 | TLT | $94.27 | $84.47 | -10.40% |
| 22 | Silver | $70.61 | $60.30 | -14.60% |
| 23 | Natural Gas | 3.514 | 2.948 | -16.11% |
| 24 | Bitcoin | $87,850 | $63,982.03 | -27.17% |
| 25 | Ethereum | $2,967 | $1,793.92 | -39.54% |
Every close here is drawn from the same price record the table itself reads from, so the numbers on this page and the numbers in our database cannot drift apart. All closes are Friday 10 July 2026. MSCI EM is derived from the EEM ETF close (66.90) multiplied by the locked index ratio (28.367); Baltic Dry is the Baltic Exchange BDI (10 Jul, 2,944). Every year-to-date figure is recomputed from the locked 1 January baselines rather than carried forward, so an error cannot compound week to week. What you can check yourself, with nothing but a calculator: every close, every baseline and every percentage in the table above. What you cannot yet check: our internal record that these were the verified closes on the day. Opening that is next.
Every company that has appeared in On the Radar is tracked here until its formal score date. A company moves from On the Radar to this appendix when there is no fresh catalyst that week, the analytical call is intact, but there is nothing new to add. Every close below is the verified close we recorded on the date shown beside it. Unlike the Scoreboard above, this table is typed, not fetched, so it carries our word rather than a live read. That is the honest distinction and you should hold us to it.
| Company | Entry | Week | Current Close | Close Date | Return | Original Thesis | Score Date |
|---|---|---|---|---|---|---|---|
| Bloom Energy (NYSE: BE) | $207.10 | Wk 14 | $244.61 | 10 Jul | +18.1% | On-site fuel-cell power for AI data centres, deployable in months versus multi-year grid queues | Thesis score 18 Jul. |
| The loud story of the week. Hunterbrook Capital published a short report alleging that Bloom is quietly dependent on Chinese scandium, a critical raw material, despite its chief executive stating since February 2025 that it has “no China supply chain”. The arithmetic in the report is the part worth reading: Bloom’s five-gigawatt production goal would need roughly 220 tonnes of scandium oxide a year, against a projected global supply of about 240 tonnes. Bloom rebutted it hard, claiming visibility to support up to 25 gigawatts. The stock fell as much as 11 percent, bounced on the rebuttal, then gave it back: it closed Friday at 244.61 dollars, down 3.8 percent on the day and plus 18.1 percent from entry, not the plus 22.8 percent it carried on Thursday. The 18 July thesis verdict now lands in the middle of a genuine fight over the company’s supply chain, and we will publish it either way. | |||||||
| Freeport-McMoRan (NYSE: FCX) | $65.49 | Wk 14 | $61.52 | 10 Jul | -6.1% | Largest listed copper pure-play on AI build-out demand; ~31% structural supply deficit versus 2035 | Thesis score 18 Jul. |
| Still below entry and behind copper, which is up double-digits while the stock is down; the miner-versus-metal gap is the whole question the 18 July verdict has to answer. | |||||||
| Talen Energy (NASDAQ: TLN) | $361.01 | Wk 15 | $385.80 | 10 Jul | +6.9% | Nuclear and gas baseload power for data centres; AWS power-purchase optionality | Thesis score 25 Jul. |
| Quietly intact, above the entry tell, the data-centre-baseload thesis not yet tested. Two weeks to the verdict. | |||||||
| Venture Global LNG (NYSE: VG) | $13.08 | Wk 16 | $12.24 | 10 Jul | -6.4% | Structural Qatar LNG gap independent of Hormuz diplomacy; highest spot exposure among US exporters | Thesis score 2 Aug. |
| A real recovery this week, from roughly minus 15 percent to minus 6.4 percent, as crude firmed back above 71 dollars and the Asian spot market steadied. The declared adverse driver, soft oil, has stopped doing damage for now. Still below entry, and the 2 August verdict is graded on the pre-committed terms, not the bounce. | |||||||
| MP Materials (NYSE: MP) | $67.21 | Wk 17 | $52.21 | 10 Jul | -22.3% | US rare-earth mine with a Defense Department cost-plus price floor; strategic-utility re-rating | Thesis score 9 Aug. |
| The loudest loser, down 22.3 percent, China rare-earth targeting the standing drag; named plainly, not buried. The thesis said Washington would re-rate MP as strategic; Beijing re-rated it first, in the other direction. | |||||||
| USA Rare Earth (NASDAQ: USAR) | $21.00 | Wk 18 | $18.48 | 10 Jul | -12.0% | Only US firm building a complete mine-to-magnet rare-earth chain; dual federal backing | 4-wk scored 13 Jun ✓; thesis 16 Aug. |
| Below entry and the weaker half of the rare-earth pair, down 12.0 percent on the Friday close. Nothing moved the thesis this week: it still rests on whether the mine-to-magnet chain gets built and funded on schedule, and it is graded against the pre-committed conditions on 16 August, not the news cycle. | |||||||
| YPF Sociedad Anónima (NYSE: YPF) | $50.31 | Wk 23 | $47.58 | 10 Jul | -5.4% | Vaca Muerta shale + ~$20bn Argentina LNG + sovereign re-rating, mispriced as a spot-oil EM cyclical | 4-wk 17 Jul; thesis ~Jun 2027. |
| Recovered from minus 11.7 percent to minus 5.4 percent as crude firmed; the four-week tactical verdict lands on 17 July with the call still behind but closing. The driver named at entry, crude beta plus Argentine country risk, has turned from adverse to supportive in the past fortnight. | |||||||
| Mitsubishi UFJ (NYSE: MUFG) | $20.17 | Wk 25 | $21.65 | 10 Jul | +7.3% | Japan rate-normalisation re-rate; repatriation flow + jumbo-hike optionality mispriced behind a weak-yen headline | 4-wk 31 Jul; thesis 3 Jul 2027. |
| Up 7.3 percent in its first week on the verified Friday close. The yen remains the adverse near-term driver for a dollar holder of the ADR while the structural rate-normalisation thesis plays out over a year. | |||||||
This week: eight active calls tracked here; Arista, the ninth, moved up to On the Radar on a fresh catalyst and is not repeated. Three of these eight are in the money (Bloom Energy, Talen, Mitsubishi UFJ) and five are behind; counting Arista, the live book is four up and five behind. The losers are reported as loudly as the winners: MP Materials, down 22.3 percent, remains the book’s biggest loser by some distance, while Venture Global and YPF both recovered materially this week as crude firmed. The first final thesis verdicts, Bloom Energy and Freeport, arrive on 18 July.
Everything below is the evidence for everything above. You do not need to read it. It is here so that you can.
| Indicator | Latest | Prior | Direction |
|---|---|---|---|
| Core PCE YoY (May 2026) | 3.4% | 3.4% | Highest since Oct 2023; the rules-based Fed’s number |
| NY Fed 1-Year Inflation Expectations (June) | 3.7% | 3.5% | A three-year high; the week’s hawkish tell |
| Nonfarm Payrolls (June 2026) | +57k | +129k (rev.) | Half of forecast; prior two months revised down 74k |
| Unemployment Rate (June 2026) | 4.2% | 4.2% | Steady; participation soft |
| ISM Manufacturing New Orders (June) | 56.0 | 56.8 | Still expanding |
| Retail Sales MoM (May 2026) | +0.9% | +0.4% | Firm; June due ~15 Jul |
| Fed Funds Rate (current) | 3.50-3.75% | 3.50-3.75% | Held; Fed Path read Hold, hawkish tilt pending 14 Jul CPI |
The front end held while the long end sold off: the whole move this week was in the term premium, the extra yield investors demand for lending for thirty years rather than two.
| Tenor | Yield | WoW Change |
|---|---|---|
| 2-Year Treasury | 4.16% | Roughly steady; the front end anchored on a soft labour read |
| 5-Year Treasury | 4.25%* | Carried; no separate verified 10 Jul print |
| 10-Year Treasury | 4.54% | +6bps |
| 30-Year Treasury | 5.05% | Cleared 5 percent; the standing red |
| 2Y-10Y Spread | +38bps | Re-steepened; positively sloped, the standing red |
| HY OAS | ~270bps | Tight |
| IG OAS | ~76bps | Stable |
*The 5-year is carried from the prior week where no separate verified print was available at production, and logged in our exceptions register and re-checked on Monday.
Oil firmed on Iran headlines while the precious metals fell; silver and natural gas led the losers.
| Commodity | Close (10 Jul) | WoW % | YTD % |
|---|---|---|---|
| WTI Crude Oil | $71.51/bbl | +4.0% | +13.2% |
| Gold | $4,128.90/oz | -1.4% | -4.9% |
| Silver | $60.30/oz | -4.0% | -14.6% |
| Copper | $6.285/lb | +1.0% | +10.6% |
| Natural Gas (Henry Hub) | $2.948/MMBtu | -9.1% | -16.1% |
| Baltic Dry Index | 2,944 | +8.4% | +56.4% |
Note: commodity closes are Friday 10 July. Baltic Dry from the Baltic Exchange BDI (10 Jul, 2,944).
| Date | Event | Relevance |
|---|---|---|
| 14 Jul 2026 | US June core CPI | The week’s tell; decides whether the extended-hold read hardens or the cut path reopens |
| Mid-Jul 2026 | Q2 earnings season, banks first | The breadth confirmation test |
| ~15 Jul 2026 | US retail sales (June) | A10 consumer-health input |
| 17 Jul 2026 | YPF 4-week tactical score date | On the Radar accountability mark |
| 18 Jul 2026 | BE & FCX thesis-horizon scores | First final dual-horizon verdicts on the book |
| 25 Jul 2026 | TLN thesis-horizon score | Final verdict on the call |
| 24 Jul 2026 | Week 28, 2026 Thesis Check-In due | Every-6th-edition Repricing dispersion mark |
| 31 Jul 2026 | MUFG 4-week tactical; ANET thesis 28 Aug | Rolling On the Radar marks |
| 2 Aug / 9 Aug 2026 | VG and MP thesis-horizon scores | Both under water; scored on pre-committed terms |
| Pair | Rate | YTD % | Driver |
|---|---|---|---|
| USD/TRY | 46.98 | +32.70% | Lira weak; domestic inflation, EM credit pressure |
| USD/ZAR | 16.346 | -6.86% | Rand firm on metals exports; YTD against the locked baseline |
| USD/JPY | ~161 | Yen near a 40-year low | BoJ at 1% vs a jumbo-hike argument; the MUFG call’s adverse driver |
| DXY (US Dollar Index) | ~99 | ~flat | Steady as the long end does the work (carried) |
| Indicator | Level | Signal |
|---|---|---|
| VIX | 15.03 | Low and stable |
| MOVE Index (bond volatility) | 72.4 | Well below the 110 caution line (Yahoo ^MOVE, 9 Jul close) |
| HY Credit Spread (OAS) | ~270bps | Below the 350bps danger zone |
| S&P % above 200dma | 67.2% | Above the 60% line (StockCharts $SPXA200R, 10 Jul) |
| 2Y-10Y Yield Spread | +38bps | Positively sloped, re-steepened, standing red |
| Insider Clusters (net selling) | 0 sectors | No net-selling cluster |
| Crash Probability Score | 15.0/100 | No Credible Crash Signal; unchanged |
Each of the seven signals scores 0 when green, 5 when amber and 10 when red. Multiply each score by its weight, add the seven together, then multiply by ten so the scale runs from 0 to 100. (If every signal were red that is 10 × 1.00 × 10 = 100, which is why the scale tops out there.) The thresholds and the weights do not change from week to week; when they do, we restate the prior week under the new rubric and say so.
| Signal | Weight | Green (0) | Amber (5) | Red (10) | This week | Score |
|---|---|---|---|---|---|---|
| High-yield credit spread (OAS) | 17.5% | < 350bps | 350–450bps | > 450bps | 270bps | 0 |
| MOVE index (bond volatility) | 17.5% | < 110 | 110–130 | > 130 | 72.4 | 0 |
| ISM new orders | 15% | > 50 | 48–50 | < 48 | 56.0 | 0 |
| Yield curve (10-year minus 2-year) | 15% | < −0.25 and stable | −0.25 to 0 | > 0 (re-steepening) | +38bps | 10 |
| VIX, level and trend | 12.5% | < 20 and stable | 20–28, or rising | > 28 | 15.0 | 0 |
| % of S&P above its 200-day average | 12.5% | > 60% | 40–60% | < 40% | 67.2% | 0 |
| Insider selling clusters | 10% | 0 sectors | 1 sector | 2 or more | 0 sectors | 0 |
The arithmetic: only the yield curve is red. It scores 10 and carries a weight of 0.15. So 10 × 0.15 = 1.5, and 1.5 × 10 = 15.0 out of 100.0 of 100. Everything else scores zero. The bands: 0–30, no credible crash signal, normal volatility expected · 30–55, elevated caution · 55–75, pre-crash conditions assembling · above 75, high crash probability. The score identifies preconditions, not outcomes: conditions can assemble and then dissipate without a crash. It tells you whether the next 90 days deserve more caution than the last 90.
Where each number comes from. High-yield spread: FRED, ICE BofA index (9 Jul). MOVE: Yahoo Finance, ^MOVE close (9 Jul). ISM new orders: the ISM June manufacturing report (released 1 Jul; it is a monthly series, so June is the current reading). Yield curve: FRED, the 10-year minus the 2-year (9 Jul). VIX: Yahoo Finance, ^VIX close (10 Jul). Percentage of the S&P above its 200-day average: StockCharts $SPXA200R (10 Jul). Insider clusters: OpenInsider, trailing four weeks. Six of the seven are free and public, and you can pull every one of them yourself.
| Company/Event | Data Point | Relevance |
|---|---|---|
| Remote Labor Index (CAIS/Scale, 1 Jul 2026) | 15.8% of end-to-end freelance projects completed, up from 2.5% at launch in Oct 2025 | More than quadrupled in under eight months (CAIS); a forward Displacer meter |
| Arista Networks (NYSE: ANET) | 8 Jul: demand for the 1.6Tbps 7060XE7 (announced 9 Jun) running ahead of plan as supply begins; Meta/Microsoft/Oracle named as first deployments; +7% on the day | A genuine new catalyst; the networking choke-point in the AI build-out (On the Radar) |
| Robotics funding | ~$18.8bn raised globally in 2026 already, past ~$15bn for all of 2025 | The fastest-moving private-capital theme; Neura, Figure, AI2, Agility |
| AI financing backlog | ~$2.1tn of remaining performance obligations across the big four platforms | Reframed as an infrastructure credit facility; Contrarian Corner’s “second derivative” |
| Flashpoint | Status | WMP Assessment |
|---|---|---|
| Tariffs as a trade-flow weapon | 25% tariff threatened on Iran’s trading partners; EU deal at a 15% cap | The lead flashpoint this week; tariffs increasingly about whether goods can move, not just their price |
| China rare-earth chokepoint | ~85% of refined rare earths still Chinese | The unresolved leverage; the channel most likely to reintroduce goods inflation |
| US-Iran / Hormuz | MoU signed 17 Jun; oil waivers revoked 7 Jul after fresh tanker attacks; tankers at ~75% of prewar; WTI $71.51 | A signature is not compliance. The residual risk is missile/production reconstitution, not a Strait closure |
| Developed-market long end | Term premium rebuilding across developed sovereigns; US 30Y >5% | The “duration is cheap” call needs the term-premium rebuild to stop first |
| Global Liquidity Cycle | Topping (Howell); ~$40tn of rollovers by 2027 | Qualifies every green credit reading; the named pressure point behind tight spreads |
Six monthly indicators of the American consumer, updated as new releases drop. One new print this week: the NY Fed’s one-year inflation expectation (June), which hit a three-year high of 3.7 percent. No high-priority flags: confidence above 85, savings above 2.5%, retail sales positive, auto sales above 15.0M. The strain lives one layer down, in expectations and the share of households worse off.
| Indicator | Current | Prior | Direction | Release |
|---|---|---|---|---|
| NY Fed 1-Year Inflation Expectations | 3.7% | 3.5% | ▲ Up to a three-year high | June 2026 (7 Jul) |
| Conference Board Consumer Confidence | 91.2 | 91.2 | Carried; next ~28 Jul | June 2026 |
| NY Fed % Worse Off Than a Year Ago | 48.0% | 48.0% | Near half of households | May 2026 (carried) |
| Retail Sales MoM | +0.9% | +0.4% | Firm; June due ~15 Jul | May 2026 (carried) |
| Auto Sales SAAR | 16.1M | 16.2M | Above the 15.0M flag line | June 2026 |
| Personal Savings Rate | 3.0% | 3.0% | Above the flag line; June PCE ~end Jul | May 2026 (carried) |
Sources: NY Fed Survey of Consumer Expectations; Conference Board; US Census Bureau; Cox Automotive / JD Power; BEA.
The Scoreboard is not typed out by hand. When this page loads, the table above fetches the week’s closes directly from our price record and draws itself from what it finds. The numbers you are reading and the numbers in our record are therefore the same numbers, by construction. One honest caveat, because we would rather tell you than be caught: a typed copy of the table also sits in this page as a fallback, and if the live fetch fails (an ad-blocker, a corporate network, a printout) you are reading that copy instead. It is generated from the same record and it matched at publication. But we cannot prove to you, in the moment, which of the two you are looking at, and we are not going to pretend otherwise. Closes come from a direct market-data feed taken after the close, never from a search, and every year-to-date figure is recomputed from the baselines we locked on 1 January rather than carried forward, so an error cannot compound week to week.
What is checkable, and what is still ours alone. Everything in the Scoreboard and the crash gauge you can check yourself with a calculator and the sources named beneath each table. What you still cannot inspect is our own audit trail, the record of which prints we accepted and which we rejected on the day. That is the last closed door in this publication, and we are working on opening it.
What is carried this week. One figure: the 5-year Treasury, where no separate verified print was available at production. It carries an asterisk wherever it appears and is logged in our exceptions register for re-checking on Monday. Two derived figures, stated plainly: MSCI EM is the EEM fund close (66.90) multiplied by the index ratio we locked in January (28.367); Baltic Dry is the Baltic Exchange reading (10 July, 2,944), cross-checked against HandyBulk (9 July, 2,910).
The Repricing line in the masthead tracks five asset classes: the S&P 500, the Bloomberg Aggregate bond index, gold, WTI crude and the high-yield credit spread. “Dispersion” is the year-to-date gap between the best and worst of the five; the split is how many are up and how many are down. At an 18-point range with three up and two down, our own measure reads the 2026 Repricing thesis as only weakly confirmed so far, and we report it that way rather than claim it is on track.
Charts and outside sources. All four charts here (the hero sparklines, the Scoreboard bars, the yield curve and the commodity moves) are drawn in our own house style from the figures above. We reproduce no third-party chart as an image. Outside research cited this week: Apollo (Torsten Slok), Groundbreaker, Neuberger Berman, Deutsche Bank Wealth, CrossBorder Capital (Michael Howell), and the Remote Labor Index. Where a source is client-only research you cannot open, we say so rather than cite it as though it were public.
The calls. Every directional call is logged at the moment it is made, at the price it was made, and scored twice: once at four weeks to test the timing, and once at a declared horizon to test the analysis. Losses are published with the same prominence as wins. On the Radar entries are what I am watching and why. They are not recommendations to buy or sell.