Read time: 10 minutes
Welcome back to The Ruck Filter.
On Tuesday the stock market put a price on something no balance sheet shows: customers who never compare. Metaâs AI agent Muse had been live for two weeks and had just passed 560,000 daily users. That was enough to take 6% off Charles Schwab and Allstate and 10% off Planet Fitness in one session, and to make telecoms Europeâs worst sector, although nobody in the EU can download the app.
Most coverage filed it under AI disruption. The better question is who collects the toll once habit stops paying, and Amazon answered it two days before the selloff.
Three sections. The inertia trade and where its rent is moving. A week in which diesel, China and Iran policy all ran on one date, and what that does to the energy chain from last week. And the long end of the Treasury curve, which has no defender, including a correction of my own. Plus a DACH box on why Germanyâs largest chemicals deal in years starts in a coal mine.
1. Signal vs. Noise: Inertia Just Got a Price đ¤
The Noise: AI agents are coming for banks, insurers and travel sites.
The Alpha: Tuesday priced the end of a moat nobody ever built. The rent that habit used to earn doesnât vanish. It moves to whoever controls access to the customer, and the fight over that access is already public.
The Filter: Meta launched Muse on 8 September, in the US only. It links to services like Gmail and OpenTable and does the work itself: bookings, price comparisons, customer-service chats, purchases once the user approves. After eleven days it had 560,000 daily users (Reuters) and had overtaken ChatGPT as the most-downloaded free iPhone app. Meta rose 11% on Monday.
On Tuesday the market went looking for the other side of that trade. The S&P 500 financials index fell 2.4%; Schwab lost more than 6%, Allstate about 6%, Planet Fitness 10.6%, Booking and Expedia close to 4%. A Goldman Sachs list of exposed companies, cited by Bloomberg, ran from AT&T and Progressive to Netflix and Uber. Bloombergâs label for what they share: consumer inertia. In the same session chip stocks rose and the Nasdaq closed at a record, with memory makers Micron and SanDisk among the gainers. The market sold the habit and bought the hardware the agents run on. In Europe, Orange and BT fell about 4% and telecoms were the Stoxx 600âs worst sector; on Thursday Scout24 hit a two-month low.
Now the part that matters for positioning. Last week I argued that the return in a supply chain accrues to the step that canât be replicated quickly. Inertia is an odd asset under that rule because nobody built it: no plant, no patent, no capex. It rested on one scarce thing, the customerâs attention, and an agent makes attention cheap. So the rent moves to the new chokepoint.
Meta has said where it wants that to be. At Connect on Wednesday, Mark Zuckerberg said Muse will stay free âfor a huge number of tokens with the expectation that over time we will profit by taking a small fee from transactions.â Amazon had already answered. Since Sunday night, Muse users trying to buy on Amazon see this: âContinued access by an unauthorized AI agent violates Amazonâs Conditions of Use.â Amazon has nothing against agents. Its own, Buy for Me, has been completing purchases on other retailersâ sites since 2025. What it refuses is someone elseâs agent standing between itself and its customer.
In Issue #026 I used a simple test for a real bottleneck: the layer a customer cannot build for himself. A customer can replace his habit with software in an afternoon. He cannot replace Amazonâs catalogue and delivery network. By that test Amazon owns a bottleneck, and much of what sold off on Tuesday was a middleman whose margin depended on nobody comparing. The dividing line is whether a company can refuse the agent. A merchant with scale can; a car insurer or a gym canât.
Schwabâs drop has a precedent. In 2023, rising rates set off âcash sorting,â clients moving idle cash from low-yield sweep deposits into money-market funds. By the third quarter Schwabâs net interest revenue was down 24% on the year, its bank deposits down 28%, the shares down about 38% year to date. That took a rate shock and twelve months. An agent that moves idle balances automatically needs neither. Reuters blamed Tuesdayâs bank selloff on the agent and on the flattest 2s10s curve since March 2025. By Friday the curve had steepened to about 36 basis points on Treasury data, yet financials still ended the week down about 2%. The curve half of the worry reversed within days. The deposit half had nothing to reverse.
A measured approach:
Sort holdings by what the customer actually pays for: a service he would choose again, or the fact that he never looked. The second is now a factor exposure, and on Tuesday it traded like one.
Watch Amazon, not the app charts. If large merchants keep agents out, the toll stays with them. If they sign on, as Walmart, Best Buy and Instacart already have, it moves to Meta. What to watch: whether Amazonâs block holds into the holidays, when Meta ships its Muse Charm device.
In Europe this is a calendar trade for now. Orange, BT and Scout24 were sold on a product their customers cannot download. What to watch: the day Meta names an EU launch date.
2. The Midterm Clock đłď¸
The Noise: Trump floated a diesel export ban, then backed off. Nothing happened.
The Alpha: Diesel, China and Iran policy all moved this week on the same timetable: keep visible prices down before 3 November, push the risk past it. That splits last weekâs energy chain in time. Until the vote, the political risk sits on the refiners that earn the margin. After it, if the Wall Street Journal is right, the war risk returns to crude. The week also exposed the variable my own rule was missing.
The Filter: On Tuesday at the UN, Trump said of an export ban, âIâve called for it,â and Treasury Secretary Bessent said the administration was examining whether a full or partial ban was feasible. The pressure came from farm-state Republicans led by Chuck Grassley, with harvest under way and AAA diesel at a record $6.53 a gallon. The $6.05 record I quoted last week is already stale. Marathon Petroleum fell 3.2% that day. On Wednesday Politico reported a plan for a 90-day ban; the White House denied it within hours. By Friday Energy Secretary Chris Wright had moved to persuasion (âYou gotta put Americans first. Reduce a little bit your exports overseasâ) and said the blunt tool of a ban âdefinitely doesnât work.â Late on Friday Ted Cruz told refining executives the White House had assured him there would be no ban. Diesel futures finished the week about 5% lower. Texas beat Iowa, for now.
The audit of my own rule. Last week I wrote that returns accrue to whichever step cannot be replicated quickly. That holds until the margin shows up at the pump in a midterm year. In late September 2022, Energy Secretary Granholm met Exxon, Marathon, Phillips 66 and Shell and floated voluntary export curbs, export limits and minimum inventories, according to Bloomberg. The industry refused; no mandate followed. Four years later a president from the other party reached for the same valve in the same season and ended with the same request. Iâd treat that as a regularity. âVoluntaryâ restraint is the soft version of the sovereign margin tax from Issues #026 and #028: the refiner is asked to sell at home below export parity.
China ran on the same clock. After Xiâs state visit, the two sides agreed reciprocal tariff cuts on about $30 billion of goods: toys, holiday decorations, small appliances and childrenâs car seats on the US side; farm goods, seafood and wood on Chinaâs, plus 10 million tonnes of US coal in each of 2027 and 2028. Holiday shoppers, farmers, coal states. I read it as a map of the electorate five weeks out.
Then Iran. Oil fell on Friday on truce hopes, November WTI settling at $92.60 and Brent near $104.75. Tehran had offered to reopen Hormuz within seven days if Washington lifted its naval blockade, waived oil sanctions and agreed a regional ceasefire. On Saturday Trump said: âI reject their proposal.â The Wall Street Journal reported he has told aides he expects a renewed bombing campaign after the midterms. Fridayâs price assumed a truce that was turned down the next morning; the risk has simply been given a later date. The physical system stays tight regardless. Aramco restarted its East-West pipeline at reduced capacity after two weeks, Yanbu loadings are still suspended (DTN), and the Houthis now hold Yemenâs Red Sea coast.
Europe sits on the other side of the valve. The US supplies about 32% of the EUâs diesel imports from outside the bloc this year, up from 17% in 2025, and about 57% in northwest Europe, per Euronews. EU Energy Commissioner Dan Jørgensen asked Trump to keep a âfree flowâ into winter; Emmanuel Macron called a ban âcatastrophic.â A US curb would invert last weekâs picture and move the crack from Texas to Europe, but only to refiners that have crude. Orlen, which draws 40% to 50% of its crude from Saudi Arabia depending on whether you take Reutersâ figure or Bloombergâs, has lost four September Aramco cargoes and bought 16 replacements. And on 16 September Reuters put physical Dated Brent near $122 while futures traded around $108. The headline crack is quoted against futures; a European refiner buying prompt barrels paid a lot more. The two October customers Aramco cut are still unnamed, and Iâm still not guessing.
A measured approach:
Treat US refining margins as politically capped until 3 November. The ban is off the table, the request isnât. What to watch: weekly EIA distillate exports, where âvoluntaryâ restraint would show up first.
For European refiners, read crude cost rather than the crack. Their margin now hangs on two governments, Riyadh for the barrel and Washington for the competing diesel. What to watch: feedstock disclosure in Octoberâs third-quarter statements.
Donât read Fridayâs oil price as de-escalation. It priced an offer rejected within a day.
3. The Long End Nobody Defends đ
The Noise: Strong data, higher yields. The bond market is pricing growth.
The Alpha: Nobody is prepared to cap long-term US borrowing costs. The Treasuryâs buybacks come with a price limit, the Fedâs chair wants a smaller balance sheet, and Wednesdayâs five-year auction found thinner demand from abroad. The markers I set in Issues #033 and #036 have both been hit. First, a correction that sharpens the point.
The Filter: Correction. In Issue #036 I wrote that the Treasury âcould not buy its full allotmentâ when it purchased $5.187 billion against a $6 billion maximum on 10 September, and I repeated it last week. I misread the mechanics. In a buyback, dealers offer bonds and the Treasury accepts only the offers it considers good value. According to The Vault Reportâs tally of Treasury results, dealers offered $10.5 billion into that operation, 1.75 times the cap. On Thursday it happened again at the long end: $10.47 billion offered, $4.08 billion accepted, 12 of 35 eligible bonds (Newsquawk). There was no shortage of sellers. The Treasury declined to pay up, and that tells you more than my original reading did. A Treasury that wonât overpay isnât trying to set the yield. The programme is liquidity support and nothing more.
The Fed isnât filling the gap. I found no Fed official referencing the Bank of Englandâs gilt approach this week, which answers the question I left open in Issue #037. The opposite happened: Kevin Warshâs plan to shrink the $6.7 trillion balance sheet has stalled, CNBC reported, with colleagues reluctant and ten-year yields above 5% making it bad timing. John Williams called another hike this year âreasonableâ; futures put about 70% on October.
Nor were buyers queueing. Wednesdayâs $70 billion five-year auction cleared at 5.033% and tailed by 3.1 basis points, with indirect bidders, the category that includes foreign accounts, down to 54.3% from 61.5% at the previous sale. On the Treasuryâs par curve the five-year closed Thursday at 5.03%, above 5% for the first time since 2007. The ten-year ended the week at 5.17% (5.01% a week earlier), the thirty-year at 5.49% (5.34%), its highest since 2004. Mortgage rates neared 7.5%; the ten-year Bund touched 3.6%, a 17-year high. And the flash US composite PMI read 58.4, the strongest since July 2021, with input prices rising at the fastest pace in four years.
In Issue #033 I said a thirty-year back near 5.30% inside the buyback window would establish the toolkitâs limits in public. In Issue #036 the marker was 5.40%. At 5.49%, both are done.
The equity derivation. The S&P 500 rose 1.2%; the equal-weighted index fell 1.1%, and seven of eleven sectors ended lower (Bull Bear Report). Technology gained 3.6%, utilities lost 4.3%, real estate 2.2%. Small caps fell 0.8%, transports 1.2%, and consumer staples dropped for a sixth straight week. In Issue #037 I called the five-week cyclical slide an energy story misread as a rates story. This week crude fell and the most rate-sensitive sectors fell hardest, so rates now belong in that diagnosis alongside fuel. The weekâs one staples report adds nuance to the volume signal I flagged: Costcoâs traffic rose 3.3% as shoppers consolidated trips. At least at Costco, volume is concentrating in the cheapest one-stop shop with a fuel station attached rather than simply disappearing.
One chain from last week didnât fire. The yen weakened to 159.00 per dollar on Thursday, short of 160. Finance Minister Satsuki Katayama said the principles of Julyâs joint intervention were âstill alive,â Bloomberg reported that she and Bessent discussed the desirability of a strong yen, and the yen closed Friday near 157.3.
A measured approach:
Price long-duration assets as if there is no backstop at these levels, because this week there wasnât one. What to watch: PCE on 30 September and payrolls on 2 October, which decide October.
Read buyback results as a gauge of the Treasuryâs price discipline, not of demand. I got that wrong once.
Track the gap between the cap-weighted and equal-weighted index. 2.3 points in one week, with the thirty-year at a 22-year high, is a narrow base.
đ DACH Desk: BASF, Evonik and the Mines Under the Ruhr
On Friday BASF confirmed exploratory talks with the RAG-Stiftung and Evonik about a possible takeover of Evonik, after reports by Dealreporter and the Financial Times. Evonik rose about 7%, BASF fell almost 4%.
The seller is the story. The RAG-Stiftung owns about 44% of Evonik and exists to fund the perpetual obligations of German hard-coal mining: pumping mine water, maintaining polders and treating groundwater, forever. In 2025 that cost âŹ309.3 million, per its accounts. The Evonik dividend it received covered about 80%. Evonik has since cut its payout to âŹ1.00, which on the foundationâs stake comes to a little over âŹ200 million, around two-thirds of that bill. The foundation had already classed 20.7% of Evonik as for sale while keeping 25.1% for the long term. Should its assets ever fall short, North Rhine-Westphalia and Saarland are liable under the 2007 inherited-liabilities contract, with the federal government covering a third. A good part of this deal is German public liability management.
What BASF would buy: a 6.1% return on capital in 2025 against Evonikâs own 11% target, in an industry that closed 37 million tonnes, about 9% of European capacity, between 2022 and 2025 while new investment dropped close to zero (Cefic). In the capital-cycle terms of Issue #027, consolidation plus closures plus no new capacity is how bottoms tend to look.
The same week, Wacker Chemie sold 2.2 million Siltronic shares for âŹ171 million, cutting its stake to about 15%, and Siltronic fell 7%. In 2022 Siltronicâs sale to GlobalWafers collapsed when Berlin failed to approve it in time; now its anchor shareholder sells in blocks. And in a founder exit, Schneider Electric offered âŹ70 a share in cash for Shelly Group, about âŹ1.2 billion in enterprise value and 27% above the undisturbed price of 28 July. Founders holding about 57% have committed to sell under conditions; closing is targeted for early 2027.
What to watch: whether BASF offers cash or shares. Cash funds the mines. Shares swap one concentration risk for another.
Outro: Defended by Default
Most of what repriced this week had been protected by something nobody chose to provide. Habit protected the margins of brokers and insurers. An open export market protected refiners until an Iowa senator noticed the harvest. Long-dated Treasuries were assumed to have a buyer of last resort, and that buyer turned out to have a price.
Protection like that goes without an announcement, which is exactly why itâs worth checking for before it does.
The Takeaway: Muse needed eleven days to reach 560,000 daily users and one session to put a price on customer habit. How much of what you own earns its margin from customers who have never compared?
Daniel Ruck
Editor, The Ruck Filter
Filter Sources this week
Reuters (via Yahoo Finance) | Muse launch, user numbers and Zuckerbergâs remarks at Meta Connect, September 23â24, 2026
Bloomberg (via Yahoo Finance) | âconsumer inertiaâ selloff, Goldman Sachs list, European telecoms, September 22, 2026
Reuters (via KELO) and Investing.com | financial-sector moves and the 2s10s curve, September 22, 2026
TechCrunch and Tech Times | Amazonâs block of Muse; Amazonâs Buy for Me agent, September 21â23, 2026
TradingKey | Nasdaq record, memory stocks, Marathon Petroleum daily moves, September 22â24, 2026
dpa-AFX (via onvista) | Scout24, September 24, 2026
InvestmentNews | Schwab third-quarter 2023 results and cash sorting
Axios; Reuters (via U.S. News, OilPrice); Benzinga | diesel export ban sequence, AAA diesel data, Wrightâs Allentown remarks, September 22â25, 2026
Bloomberg; The Hill | Senator Cruzâs assurances; the Republican split over the ban, September 2026
Bloomberg (via EnergyNow) | Granholmâs meeting with refiners, September 2022
DTN | NYMEX and ICE settlements, Saudi pipeline status, September 18â25, 2026
Axios | USâChina tariff agreement (White House fact sheet), September 26, 2026
AP (via PBS) and Al Jazeera | Iranâs seven-day proposal, Trumpâs rejection, Wall Street Journal reporting on post-midterm plans, September 25â26, 2026
Euronews; Washington Examiner | EU diesel import dependence, Macron and Commissioner Jørgensen, September 2026
Bloomberg (via EnergyConnects) and Reuters (via Hydrocarbon Processing) | Orlenâs supply position and Dated Brent, September 16â18, 2026
U.S. Department of the Treasury | Daily Treasury Par Yield Curve Rates, September 2026
The Vault Report and Newsquawk | Treasury buyback operation results, September 10 and 24, 2026
CNBC; Reuters (via Investing.com) | Warshâs balance-sheet plans, thirty-year yield, John Williams, September 24â25, 2026
Bloomberg (via Edward Conard); investingLive; TFTC | five-year yield and auction statistics, September 23, 2026
S&P Global | flash US PMI, September 23, 2026; dpa (via ms-aktuell) â Bund yields, September 25, 2026
Bull Bear Report and Palma Futures | weekly index and sector performance, September 25â27, 2026
Costco | fourth-quarter fiscal 2026 results and earnings call, September 24, 2026
FXStreet, Bloomberg, Investrade | yen levels, Katayama and Bessent, September 24â25, 2026
BASF press release, September 25, 2026; Financial Times and Dealreporter (via dpa-AFX)
RAG-Stiftung | annual accounts 2025; Evonik â full-year 2025 results; Cefic â Closures & Investments Radar, January 2026
dpa-AFX (via finanzen.at, onvista) â Wacker/Siltronic placement and Shelly offer, September 22â24, 2026
Disclaimer: The Ruck Filter is for informational purposes only and does not constitute financial, investment, or tax advice. The information provided is based on data available at the time of writing and is subject to change. Investing in financial markets involves risks, including the potential loss of principal. Every reader is solely responsible for their own trading and investment decisions. Please conduct your own due diligence or consult with a licensed professional before making any financial commitments. Companies with potential conflicts of interest are excluded from coverage as a matter of editorial policy.